Domain adaptation has recently been widely explored for 3D detection. Previous works mainly use unsupervised domain adaptation (UDA) to address domain discrepancies. Despite notable improvements, their performance still largely trails models trained with fully annotated target data, due to larger domain gaps caused by different sensors and changing environments. In this paper, we exploit key characteristics of autonomous driving scenarios, including similar scenes and classimbalanced distributions, and explore a new task named active domain adaptation (ADA) for 3D object detection, which selects partial but important target data for annotation to further improve target-domain performance. Such a setting better reflects practical deployment in practice, where annotating all target-domain point clouds is prohibitively expensive while limited labels can substantially guide adaptation effectively. To this end, we propose a hybrid bi-domain active learning strategy, Bi3D++, to sample valuable data from both source and target domains and transfer source-domain knowledge to the target domain. Bi3D++ first samples target-like source data by measuring scene-level and instance-level similarity between domains, avoiding interference from irrelevant source data. Then, a hybrid active target sampling strategy selects target data by jointly considering rare-class similarity, intra-frame diversity, and inter-frame diversity, enabling diverse frames with diverse instances while emphasizing rare classes. Experiments on multiple cross-domain settings, including cross-beam and cross-location, show that Bi3D++ outperforms state-of-theart UDA methods with only 1% labeled target data and consistently improves performance as target annotations increase.

Synthetic Aperture Radar (SAR) imagery provides all-weather, all-day, and high-resolution imaging capabilities but its unique imaging mechanism abstract imagery that severely lacks the high-fidelity contours and textures essential for automated interpretation and demanding pixel-level downstream tasks. Translating SAR images into optical images is a promising solution to enhance interpretation and support downstream tasks. Most existing research focuses on scene-level translation, with limited work on object-level translation due to the scarcity of paired data and the challenge of accurately preserving contour and texture details. To address these issues, this study proposes a keypoint-guided diffusion model KeypointDiff for SAR-to-optical image translation of unpaired aircraft targets. leverages keypoints as modality-agnostic structural anchors, enabling a novel training strategy that establishes structural-level correspondence between the unpaired SAR and optical domains. Based on the classifier-free guidance diffusion architecture, a class-angle guidance module (CAGM) is designed to integrate class and angle information into the diffusion generation process. Furthermore, a detector-based supervision loss and a visual consistency loss are employed to improve image fidelity and detail quality, tailored for aircraft targets. During sampling, aided by a pre-trained keypoint detector, the model eliminates the requirement for manually labeled class and azimuth information, enabling automated SAR-to-optical translation. Experimental results demonstrate that the proposed method outperforms existing approaches across multiple metrics, providing an efficient and effective solution for object-level SAR-to-optical translation and pixel-level detail recovery. Moreover, the method exhibits strong zero-shot generalization to untrained aircraft types, highlighting the model's practical applicability.

Iterative-based methods have become mainstream in stereo matching due to their high performance. However, these methods heavily rely on labeled data and face challenges with unlabeled real-world data. To this end, we propose a consistency-aware self-training framework for iterative-based stereo matching for the first time, leveraging real-world unlabeled data in a teacher-student manner. We first observe that regions with larger errors tend to exhibit more pronounced oscillation characteristics during model prediction. Based on this, we introduce a novel consistency-aware soft filtering module to evaluate the reliability of teacher-predicted pseudo-labels, which consists of a multi-resolution prediction consistency filter and an iterative prediction consistency filter to assess the prediction fluctuations of multiple resolutions and iterative optimization respectively. Further, we introduce a consistency-aware soft-weighted loss to adjust the weight of pseudo-labels accordingly, relieving the error accumulation and performance degradation problem due to incorrect pseudo-labels. Extensive experiments demonstrate that our method can improve the performance of various iterative-based stereo matching approaches in various scenarios. In particular, our method can achieve further enhancements over the current SOTA methods on several benchmark datasets.

Bimanual manipulation is crucial in robotics, enabling complex tasks in industrial automation and household services. However, it poses significant challenges due to the high-dimensional action space and intricate coordination requirements. While video prediction has been recently studied for representation learning and control, leveraging its ability to capture rich dynamic and behavioral information, its potential for enhancing bimanual coordination remains underexplored. To bridge this gap, we propose a unified diffusion-based framework for the joint optimization of video and action prediction. Specifically, we propose a multi-frame latent prediction strategy that encodes future states in a compressed latent space, preserving task-relevant features. Furthermore, we introduce a unidirectional attention mechanism where video prediction is conditioned on the action, while action prediction remains independent of video prediction. This design allows us to omit video prediction during inference, significantly enhancing efficiency. Experiments on two simulated benchmarks and a real-world setting demonstrate a significant improvement in the success rate over the strong baseline ACT using our method, achieving a \textbf{24.9\%} increase on ALOHA, an \textbf{11.1\%} increase on RoboTwin, and a \textbf{32.5\%} increase in real-world experiments. Our models and code are publicly available at https://github.com/return-sleep/Diffusion_based_imaginative_Coordination.

In the Tianwen-1 Mars exploration mission, ultrawideband radar is carried by the Zhurong Martian rover. The exponential attenuation at the center frequency was applied to invert the loss tangent of Mars regolith in previous studies. Ignoring the frequency-dependent absorption in the ultrawideband might cause a large error in the inversion results. Considering the transmitted linear frequency-modulated (LFM) waves, in this letter, an analytical formula for the attenuation of ultrawideband waves to invert the loss tangents of Martian regolith is derived with the accumulation of the frequency-dependent attenuated spectrum. The newly inverted loss tangent is much larger than the inverted values with the center frequency. In addition, the inversion of the loss tangent from the ultrawideband radar data obtained in the Chang’e-5 lunar program is discussed. This letter presents a corrected inversion method for the loss tangent of regolith from ultrawideband radar echoes.

Current synthetic aperture radar (SAR) missions with short repeat times bring the opportunity to get large-scale deformations in near real-time. Since nonstationary deformation is more likely to be a risk, postclassification is widely used to identify the temporal deformation patterns within extensive interferometric SAR (InSAR) data. However, conventional 3-D phase unwrapping (PU) with the assumption of stationary deformation behavior has a high probability of unwrapping error over a long time series, leading to more false detected anomalies. Here, we proposed a recursive 3-D PU method to unwrap the 3-D data stack recursively and detect the deformation anomalies concerning the nonstationary deformation. This method involves recursive temporal PU with a temporal smoothness constraint, followed by iterative spatial PU. Then, the multiple hypothesis test (MHT) is used to determine the optimal deformation model. Scatterers with deformation anomalies are identified using a generalized ratio test and assessed by their detectability power based on the predicted phase residuals. The main advantage of the proposed algorithm is the capability of dynamic data processing and decreasing the false alarm in detected deformation anomalies. The experimental results by both the simulated and real data demonstrate that the proposed method achieves reliable 3-D PU concerning nonstationary deformation, which would be beneficial for near real-time evaluation of deformation risks.

Three-dimensional target reconstruction from synthetic aperture radar (SAR) imagery is crucial for interpreting complex scattering information in SAR data. However, the intricate electromagnetic scattering mechanisms inherent to SAR imaging pose significant reconstruction challenges. Inspired by the remarkable success of 3D Gaussian Splatting (3D-GS) in optical domain reconstruction, this paper presents a novel SAR Differentiable Gaussian Splatting Rasterizer (SDGR), specifically designed for SAR target reconstruction. Our approach combines Gaussian splatting with the Mapping and Projection Algorithm to compute scattering intensities of Gaussian primitives and generate simulated SAR images through SDGR. Subsequently, the loss function between the rendered image and the ground truth image is computed to optimize the Gaussian primitive parameters representing the scene, while a custom CUDA gradient flow is employed to replace automatic differentiation for accelerated gradient computation. Through experiments involving the rendering of simplified architectural targets and SAR images of multiple vehicle targets, we validate the imaging rationality of SDGR on simulated SAR imagery. Furthermore, the effectiveness of our method for target reconstruction is demonstrated on both simulated and real-world datasets containing multiple vehicle targets, with quantitative evaluations conducted to assess its reconstruction performance. Experimental results indicate that our approach can effectively reconstruct the geometric structures and scattering properties of targets, thereby providing a novel solution for 3D reconstruction in the field of SAR imaging.

This study presents a novel method for learning terrain scattering model parameters by leveraging massive multi-source Earth observation data, aiming to achieve realistic Synthetic Aperture Radar (SAR) data simulation. By integrating Gaofen-3 and Sentinel-1 SAR data with auxiliary datasets, the scattering characteristics of various terrains were extracted and analyzed with respect to angle, season, and resolution. For forward modeling, the scattering models were compared to identify suitable models and parameters. To address the challenge of multiple solutions in parameter learning, multi-angle scattering characteristics were employed for initial value estimation, supported by a probability density-based loss function. During parameter learning, targeted learning rates were set based on the gradients of the scattering models with respect to the parameters. Extensive evaluations demonstrate that the proposed method reliably estimates scene parameter maps while preserving texture features, with simulated multi-angle SAR data based on these maps showing good radiometric consistency with measured data. This work holds considerable application potential, and integrating it with other multi-source data and neural networks will yield more valuable outcomes in the future.

Deep learning models in satellite onboard enable real-time interpretation of remote sensing images, reducing the need for data transmission to the ground and conserving communication resources. As satellite numbers and observation frequencies increase, the demand for satellite onboard real-time image interpretation grows, highlighting the expanding importance and development of this technology. However, updating the extensive parameters of models deployed on the satellites for spaceborne object detection model is challenging due to the limitations of uplink bandwidth in wireless satellite communications. To address this issue, this paper proposes a method based on parameter-efficient fine-tuning technology with low-rank adaptation module. It involves training low-rank matrix parameters and integrating them with the original model’s weight matrix through multiplication and summation, thereby fine-tuning the model parameters to adapt to new data distributions with minimal weight updates. The proposed method combines parameter-efficient fine-tuning with full fine-tuning in the parameter update strategy of the oriented object detection algorithm architecture. This strategy enables model performance improvements close to full fine-tuning effects with minimal parameter updates. In addition, low rank approximation is conducted to explore intrinsic dimensions of parameter matrices in normal-size models. Extensive experiments conducted on the DOTAv1.0, HRSC2016, and DIOR-R datasets verify the effectiveness of the proposed method. By fine-tuning and updating only 12.4% of the model’s total parameters, it is able to achieve 97% to 100% of the performance of full fine-tuning models. Additionally, the reduced number of trainable parameters accelerates model training iterations and enhances the generalization and robustness of the oriented object detection model. The source code is available at: https://github.com/fudanxu/LoRA-Det.

Synthetic aperture radar (SAR) altimeter is able to measure height with high precision, which has been extensively applied to airborne aircrafts for positioning. With the assumption of a flat surface following a Gaussian distribution, the height is obtained by retracking the waveform. However, this assumption often fails in a complex terrain, leading to unpredictable height bias. In this article, a slope-based height correction (SHC) toward robust height inversion in complex terrain is proposed using linear terrain decomposition. Based on the radar propagation equation, the impact of the topography on height inversion is investigated. Then, the response from complex topography can be divided into a determined part related to a set of primary slopes and a stochastic part with certain undulation. Additionally, the height bias induced by the slopes is given based on power constraints. The error bound of the height correction is also derived correspondingly. The main advantage of the proposed method is reliable height correction with the prior digital elevation model (DEM). Experimental results based on both simulated and real data demonstrate a significant decrease in height bias, which greatly extends the application of the altimeter.

This study introduces a novel data assimilation framework, Dynamic Compressed Sensing-Dynamic Mode Decomposition(DCS-DMD), for real-time global four-dimensional(4-D) ionospheric electron density imaging and short-term prediction. Unlike traditional methods relying on complex predefined models, the framework employs a Koopman-based algorithm to extract time-varying ionospheric features and integrate them with observational data, enabling simplified and effective ionospheric imaging and prediction. Applied to the May 10–11, 2024 geomagnetic storm, the DCS-DMD model—using Global Navigation Satellite System(GNSS) and Radio Occultation (RO) data at a 5-minute resolution—outperforms existing models in tomographic accuracy. It shows significant improvement in differential slant total electron content(dSTEC) evaluations across reference stations at various latitudes, particularly when combining GNSS and RO data. The model also aligns well with ionosonde measurements, even during geomagnetic storms, detecting a density enhancement at the storm’s onset and a suppression during the recovery phase. Furthermore, the framework demonstrates effective short-term electron density prediction, highlighting its potential for forecasting foF2 in shortwave communication. This model significantly enhances the accuracy of ionospheric imaging and forecasting, providing a streamlined tool for space weather monitoring.

Three-dimensional shape retrieval from synthetic aperture radar (SAR) images has long presented a significant research challenge, with single-view reconstruction proving even more complex due to constraints such as the scarcity of labeled data, limited sample diversity, and heightened sensitivity to radar scattering characteristics. Recently developed deep learning-based methods have made progress in single-view target reconstruction from SAR images. However, these methods still rely heavily on 3-D ground-truth supervision and fail to fully leverage SAR imaging mechanisms for 3-D reconstruction. To address these limitations, an end-to-end unsupervised single-view 3-D reconstruction framework based on a differentiable SAR renderer (DSR) is proposed, achieving precise reconstruction while eliminating the need for ground-truth data. Specifically, the encoder-decoder architecture effectively extracts 3-D and angular features, utilizing template deformation to preserve both fine details and global structures across scales, along with essential pose information for 3-D shape reconstruction. The reconstructed 3-D model is projected onto a 2-D plane, and pixel-level intersection over union (PIoU) loss is employed for unsupervised learning, enabling the extraction of discriminative latent structures and patterns. This approach effectively reduces low-frequency noise, sharpens critical edges, and enhances high-frequency details, improving spatial structure accuracy while minimizing shape distortions and height errors in complex targets. Extensive quantitative and qualitative experiments on both simulated and real datasets demonstrate the framework’s superior performance in single-view SAR 3-D target reconstruction, offering a promising solution with broad potential applications.

Researchers in the fields of ecological environment and remote sensing pay considerable attention to the estimation of forest canopy height with synthetic aperture radar (SAR). The interest is due to the ability of SAR to penetrate the forest canopy and its sensitivity to forest properties through backscattered intensity. Recent advances in deep learning (DL) present the possibility to derive canopy height maps from single high-resolution (HR) SAR images using neural networks. SAR2Canopy, an innovative framework, is proposed in this paper for canopy height estimation based, which incorporates sensor and scattering knowledge into the estimation. The integration framework allows for the merging of reconstruction trunk scattering features by an equivalent dihedral corner reflector (DCR) scattering model into the supervised tree height estimation process. The proposed method attempts a new approach combining the physical characteristics of SAR data with the nonlinear feature learning ability of DL, potentially extending to different DL algorithms. Experiments are conducted with airborne fully polarimetric P-band SAR data from two areas. Compared to the baseline models, including UNet, DeepLabV3_ResNet50, and FCN_ResNet50, the proposed SAR2Canopy integrated with the DCR model achieves an increase of up to 13.66% in the coefficient of determination (R2), while reducing the root mean squared error (RMSE) and mean absolute error (MAE) by as much as 14.59% and 20.69%, respectively.

Synthetic aperture radar (SAR) image simulation has attracted much attention due to its great potential to supplement the scarce training data for deep learning algorithms. Consequently, evaluating the quality of the simulated SAR image is crucial for practical applications. The current literature primarily uses image quality assessment (IQA) techniques for evaluation that rely on human observers' perceptions. However, because of the unique imaging mechanism of SAR, these techniques may produce evaluation results that are not entirely valid. The distribution inconsistency between real and simulated data is the main obstacle that influences the utility of simulated SAR images. To this end, we propose a novel trustworthy utility evaluation framework with a counterfactual explanation for simulated SAR images for the first time, denoted as X-Fake. It unifies a probabilistic evaluator and a causal explainer to achieve a trustworthy utility assessment. We construct the evaluator using a probabilistic Bayesian deep model to learn the posterior distribution, conditioned on real data. Quantitatively, the predicted uncertainty of simulated data can reflect the distribution discrepancy. We build the causal explainer with an introspective variational auto-encoder (IntroVAE) to generate high-resolution counterfactuals. The latent code of IntroVAE is finally optimized with evaluation indicators and prior information to generate the counterfactual explanation, thus revealing the inauthentic details of simulated data explicitly. The proposed framework is validated on four simulated SAR image datasets obtained from electromagnetic models and generative artificial intelligence approaches. The results demonstrate the proposed X-Fake framework outperforms other IQA methods in terms of utility. Furthermore, the results illustrate that the generated counterfactual explanations are trustworthy, and can further improve the data utility in applications.

Synthetic Aperture Radar (SAR) offers all-weather, high-resolution imaging capabilities, but its complex imaging mechanism often poses challenges for interpretation. In response to these limitations, this paper introduces an innovative generative model designed to transform SAR images into more intelligible optical images, thereby enhancing the interpretability of SAR images. Specifically, our model backbone is based on the recent diffusion models, which have powerful generative capabilities. We employ SAR images as conditional guides in the sampling process and integrate color supervision to counteract color shift issues effectively. We conducted experiments on the SEN12 dataset and employed quantitative evaluations using peak signal-to-noise ratio, structural similarity, and fréchet inception distance. The results demonstrate that our model not only surpasses previous methods in quantitative assessments but also significantly enhances the visual quality of the generated images.

SAR image super resolution algorithms have been shown to significantly improve performance in many applications. For these memory-intensive algorithms, large-scale images must be divided into overlapped blocks and finally concatenated back to the final result. Orthogonal subspace decomposition method has better performance, but needs to manually choose the size of noise subspace according to the noise levels. Unsuitabe size of noise subspace results in the block effect. In this work, a relative optimal noise subspace determination algorithm is proposed. A Laplace operator is applied to get the size of the relative optimal noise subspace. Considering large-scale SAR images, a chip-merging approach is designed to eliminate the blocking effect with the relative optimal size of noise subspace. Experiment shows that this algorithm eliminates the blocking effect. This work can also be applied to other methods that need the parameter selection based on data.

SAR Tomograpgy is a standard tool for 3D radar imaging, which overcomes the limitation of SAR 2D geometric distortion by using the baseline diversity. Conventional multi-baseline SAR data by repeat-pass space-borne SAR mission requires a long waiting time. The new airborne array InSAR system can acquire the multi-baseline images in a single flight, which significantly improves the practical capability of SAR based 3-D reconstruction. However, the array InSAR system with many channels has very high complexity. The performance of existing algorithms decrease significantly with such a sparse acquisition. In this paper, we have proposed a Range-Elevation (R-E) combined TomoSAR algorithm aiming to solve this issue. The conventional TomoSAR imaging model is converted to a 2D spectrum estimation. The neighboring pixels in the range domain are jointly processed to get the final height estimation. The experiments using real array-InSAR data show that R-E TomoSAR has a good performance in 3-D reconstruction with few acquisitions.

Synthetic Aperture Radar (SAR) possesses all-weather, all-time capabilities, but SAR imagery is highly sensitive to the observation configuration, with images from different viewing angles exhibiting significant variations. This poses challenges for the task of detection and recognition in SAR imagery. Inspired by the concept of neural radiance fields, this paper combines SAR imaging principles with neural networks. Our aim is to generate SAR images from other observation angles using a limited set of observation angles, thereby enhancing the physical interpretability of SAR neural networks. Additionally, within the SAR-NeRF framework, we introduce a multi-resolution hash encoding method. This approach divides the SAR imaging space into a multi-resolution grid, enhancing the training speed of SAR-NeRF and reducing the model size. We conducted experiments on the MSTAR dataset and performed quantitative evaluations using Peak Signal-to-Noise Ratio. The results indicate that our method not only matches or surpasses previous methods in generative performance but also boasts shorter training times and smaller model sizes, demonstrating its potential applicability.

This study proposes a high-dimensional characterization framework for multi-scale natural objects and surfaces scattering properties. The framework employs massive multisource remote sensing data to effectively extract the polarization scattering properties of different objects at multiple global locations. In this research, scattering models such as the Small Perturbation Method (SPM), Integral Equation Model (IEM), and Vector Radiative Transfer (VRT) were used for precise fitting of scattering properties. Utilizing these fitting results as initial values allows for stable and efficient inversion of the texture parameters of the scene. Based on texture parameters, Synthetic Aperture Radar (SAR) images under multiple observation angles were generated, and the comparison with measurement data demonstrates high R consistency and Pearson correlation.

Synthetic aperture radar (SAR) altimetry has the capability of altimetry with a good accuracy, especially for the ocean scene. The model-based methods are widely used in the height inversion of the altimetry data due to the strong correlation with physical process. However, the bias of the model will significantly decrease performance of height inversion. In this work, we introduce a novel height inversion algorithm based on the closed-form expression, which is an effective model for measuring radar echoes. Effective initialization strategy expands the application scope, and two-step retracking algorithm is designed which could be used in an iterative least-squares algorithm. In essence, this method aims to remove the coupling between the parameters in closed-form expression. Experimental results using the simulated data indicate that this proposed method has wider application range under different signal-to-noise condition. In addition, precision of two-step method remains stable compared with traditional method, which has potential for actual altimetry system.

The electromagnetic inverse task has long been recognized as a challenging research problem, which attracted substantial attention from the microwave community. In this paper, our objective is to explore the intricate relationship between geometric model imaging and radar view angles in Synthetic Aperture Radar (SAR) images, mainly focusing on the inverse problem of radar view angle estimation given a target model. However, the high cost and limited availability of SAR data acquisition, along with background interference and complex imaging mechanisms in SAR images, present significant challenges to the generalization, robustness, and accurate feature extraction of existing methods. To address these issues, we propose an interactive deep reinforcement learning (DRL) framework, which facilitates the interaction between the agent and an embedded electromagnetic simulator environment to simulate a human-like process of angle prediction step-by-step. A large number of experimental results verified that the proposed method can accurately predict the radar perspective of SAR images.

Recovering the target three-dimensional (3D) information from Synthetic Aperture Radar (SAR) images and then reconstructing have always been a challenging research topic. Inspired by computer graphics and differential geometry theory, this paper proposes a forward and inverse integration method for accurate estimation of 3D geometric parameters. A differentiable ray tracing engine is developed for forward and inverse reconstruction from SAR images, referred to as DRT. The approach derives geometry parameter gradients based on SAR scattering and mapping projection imaging mechanisms. The difference between the reference SAR image and the simulated image as the objective function to generate high-quality 3D geometric parameter recovery. This method solves the non-differentiable problem of visibility in the ray tracing process. The effectiveness of the method is verified through simulation experiments. According to the optimized 3D mesh parameters, SAR images from other observation angles can be generated. Therefore, in addition to recovering target information and reconstruction from SAR images, DRT can also be used for multi-view sample expansion tasks.

The inversion of forest structural parameters contributes to evaluation biodiversity and ecosystem functions, as well as enabling sustainable forest management. In this study, we propose a novel method for extracting forest parameters adopting P-band polarimetric synthetic aperture radar (SAR) data. Firstly, we utilize the Freeman-Durden decomposition to extract double-bounce scattering information, including ground-scatterer scattering, scatterer-ground scattering. Then, the scattering amplitude of tree trunks based on the principles of coherent scattering modeling is calculated. Additionally, we introduce the finite cylinder scattering amplitude function and utilize the constraints of the allometric growth model of vegetation to invert forest parameters. The reliability and effectiveness of the proposed method are verified through measurement data, with an RMSE of 3.35m for the inversion results. This research provides a new approach for inverting forest parameters and has potential applications in forest monitoring and ecological studies

The paper proposes an optimal sensing principle for synthetic aperture radar (SAR) imaging, maximizing the mutual information between the sensed object and the reconstructed image with the optimal SAR measurement matrix. Inspired by Shannon’s capacity theorem, the 2-D SAR sensing capacity is derived, which represents the maximum mutual information that can be acquired per unit area in 2-D scenarios. The SAR sensing capacity serves as a theoretical performance bound, guiding the design of SAR sensing systems and enabling reasonable estimation of systems’ performance. Furthermore, the optimal sensing principle is applied to the variable-resolution SAR (VR-SAR) imaging system. An equivalent experiment With Sentinel 1 Raw Data is conducted to verify the advantages of VR-SAR based on the optimal sensing principle.

Synthetic Aperture Radar (SAR) has been widely applied due to its capability for continuous monitoring under all-weather and all-time conditions. Unlike optical images, targets in SAR images are highly sensitive to observation angles whose shapes and shadows change significantly under different viewing angles. Some current methods tackle the limited angle coverage problem in SAR targets through multi-view sample generation. However, for SAR image similarity assessment, existing techniques mainly concentrate on category and scene similarity, with insufficient attention to viewing angle similarity. This paper proposes a novel method for evaluating target viewing angle similarity in SAR images. Specifically, the method introduces a Siamese network to map SAR target features from high-dimensional space to a low-dimensional manifold. A contrastive loss function is constructed using Euclidean distance to measure the similarity of low-dimensional features of viewing angles. Additionally, a sample angle pairing method is designed for model training. Furthermore, a recognition module is incorporated into the model to assess the similarity of categories. Experiments based on the MSTAR dataset demonstrate the efficiency of the proposed method in evaluating SAR target viewing angle and category similarity.

Synthetic aperture radar (SAR) offers all-weather and all-day high-resolution imaging, yet its unique imaging mechanism often necessitates expert interpretation, limiting its broader applicability. Addressing this challenge, this letter proposes a generative model that bridges SAR and optical imaging, facilitating the conversion of SAR images into more human-recognizable optical aerial images. This assists in the interpretation of SAR data, making it easier to recognize. Specifically, our model backbone is based on the recent diffusion models, which have powerful generative capabilities. We have innovatively tailored the diffusion model framework, incorporating SAR images as conditional constraints in the sampling process. This adaptation enables the effective translation from SAR to optical images. We conduct experiments on the satellite GF3 and SEN12 datasets and use structural similarity (SSIM) and Fréchet inception distance (FID) for quantitative evaluation. The results show that our model not only surpasses previous methods in quantitative evaluation but also significantly improves the visual quality of the generated images. This advancement underscores the model’s potential to enhance SAR image interpretation.

Accurate global-scale ionospheric electron density modeling is crucial for space weather monitoring, exploration, and radio signal applications. This letter presents a novel global-scale ionospheric tomography modeling method, dynamic compressed sensing-principal component analysis (DCS-PCA), building upon the previous region method CS-PCA. The upgraded method operates globally, utilizing dynamic data-driven techniques and undifferenced observation data processing to achieve high-precision quasi-real-time global-scale ionospheric tomography based on global navigation satellite system (GNSS) data. Tomographic models with a 5-min temporal resolution were constructed in this study, utilizing data from various IGS ground stations worldwide and employing the U-DCS-PCA, D-DCS-PCA, and traditional constrained algebraic reconstruction technique (CART). The DCS-PCA model is found to outperform both the CART model and the CODE Global Ionospheric Maps (CODG) model. Specifically, when evaluating differential STEC (dSTEC) errors at independent reference stations across various latitudes, we observed that the error of the DCS-PCA model is not significantly impacted by station sparsity, consistently remaining lower than that of CODG products. In contrast, the error of the CART model increases as the number of stations decreases. Additionally, the U-DCS-PCA model is found to closely align with electron density observations from ionosonde stations. This method is ideal for global 4-D ionospheric monitoring and has potential applications in space weather monitoring, exploration, and radio signal enhancement.

The ionosphere is a vitally dynamic charged particle region in the Earth’s upper atmosphere, playing a crucial role in applications such as radio communication and satellite navigation. The slant total electron contents (STECs) are an important parameter for characterizing wave propagation, representing the integrated electron density along the ray of radio signals passing through the ionosphere. The accurate prediction of STEC is essential for mitigating the ionospheric impact particularly on Global Navigation Satellite Systems (GNSS). In this work, we propose a high-precision STEC prediction model named deep neural operator network (DeepONet)-STEC, which learns nonlinear operators to predict the 4-D temporal-spatial integrated parameter for the specified satellite-ground station ray path globally. As a demonstration, we validate the performance of the model based on GNSS observation data for global and US Continuously Operating Reference Stations (CORS) regimes under ionospheric quiet and storm conditions. The DeepONet-STEC model results show that the three-day 72 h prediction in quiet periods could achieve high accuracy using observation data by the precise point positioning (PPP) with temporal resolution$30~\rm {s}$. Under active solar magnetic storm periods, the DeepONet-STEC also demonstrated its robustness and superiority than traditional deep learning methods. This work presents a neural operator regression architecture for predicting the 4-D spatiotemporal ionospheric state for satellite navigation system performance, which may be further extended for various space applications and beyond.

The accuracy of ionospheric delay modeling for user stations is intimately tied to the precise characterization of the ionospheric information in the domain of Global Navigation Satellite Systems (GNSSs). Current methods for model identification often face difficulties due to the scarcity of data from limited and sparsely located ground reference stations, and the irregular ionospheric characteristics during active periods. This is particularly true in active low latitudes, where disturbances, including GNSS signal scintillation and influence outcomes. This article introduces a universal framework, termed the global and local consistency methodology (GLCM), dedicated to extracting ionospheric spatial information by aligning estimated characteristics across global and subset spatial information. The proposed model adheres to a specifically designed objective to generate the appropriate form of functions and, based on them, to derive the ionospheric information for given areas. We carried out the simulation test to intuitively demonstrate the capabilities to improve the accuracy of the model in a direct and noninterference way. In addition, the model has been verified based on real-world data at low latitudes from a network of ground GNSS stations from all visible Global Position System (GPS) and GALILEO (GAL) satellites. The model achieves a reduction in the root-mean-square error (RMSE) of differential slant total electron content (dSTEC) by approximately 18% and 15% compared with the multiquadratic model and the Kriging model, respectively, during periods of high ionospheric activity. The proposed model has demonstrated effectiveness in ionospheric modeling and is actively being adapted for a wide range of GNSS applications and beyond.

With the rapid development and widespread deployment of radar technology, the interpretation of the vast amount of synthetic aperture radar (SAR) imagery obtained daily has emerged as a hot topic. The complexity of the electromagnetic scattering mechanisms contained within radar images makes SAR image interpretation a challenging task. Current methodologies for SAR image interpretation primarily involve feature extraction-based techniques, categorized into image-domain and frequency-domain algorithms. However, these methods are faced with issues, such as rough segmentation in images, high-computational complexity, and poor robustness, presenting significant challenges in the field. In this article, a novel framework for SAR image interpretation is proposed, leveraging reinforcement learning (RL) for the extraction of electromagnetic scattering features and the inversion of parameters. Within this framework, a nonsparse reward function, combined with curriculum learning, is introduced as the supervisory information. It enables more efficient policy updates through a structured two-stage training approach. In addition, an algorithm that integrates a four-neighbor breadth-first search (BFS) with the watershed segmentation process is proposed, aiming to enhance the accuracy of scattering center analysis in SAR imagery. The attribute scattering center model (ASCM) is utilized as a prototype for conducting algorithmic research and experimentation. Experimental results on both simulation data and measured data have indicated that the proposed method significantly improves efficiency while ensuring accuracy, demonstrating its capability to extract parameters from measured data in most scenarios.

Spaceborne multitemporal interferometric synthetic aperture radar (MT-InSAR) technique is widely used in mapping large-scale deformation with high precision. As the time series lengthens, radar scatterers with a surface change will suffer from a loss of coherence, denoted as the temporary coherent scatterers (TCSs), which can be identified by analyzing the amplitude time series. However, current amplitude analysis is sensitive to noise, especially for short temporal subsets, which leads to a high probability of false detected candidates. In this article, an amplitude-aided 3-D phase unwrapping (PU) is proposed to achieve a reliable TCS detection. First, the pre-selection of TCS candidates is conducted using two hypothesis tests. Then, an adaptive thresholding approach is proposed to get the relative optimal thresholds for varying lengths of time series. Based on the initial step-times, a hybrid 3-D PU algorithm is developed to jointly separate the noise from the candidates and refine the moments of the step-times. Finally, three application-based taxonomies are given to fuse the TCS temporal subsets and show their different temporal patterns. Experimental results using real TerraSAR-X and COSMO-SkyMed images show that the proposed amplitude-aided 3-D PU has a low probability of false alarming and an increase in the ensemble coherence. The use of different types of TCS would be beneficial to both deformation monitoring and urban change detection.

Synthetic aperture radar (SAR) has emerged as a vital tool for ship monitoring due to its all-weather, all-day high-resolution imaging capabilities. In practical operations, the wide coverage and sparse ship distribution in very large full-scene SAR images pose challenges in terms of low efficiency and high false alarm rates. Traditional methods perform poorly in complex scenarios, while deep learning (DL) methods have high computation cost. This study proposes a fast progressive detection algorithm for ship targets in large SAR images, combining the advantages of traditional Non-DL methods and DL approaches. First, at a global scale, image preprocessing operations based on traditional methods are designed to quickly extract candidate regions. Then, at regional scale, an oriented ship detector is designed for refined ship detection within candidate regions. Finally, at individual-target scale, a false alarm discrimination network is constructed to further remove false alarms. Experimental results on GF-3 full-scene SAR images demonstrate that the proposed method can achieve minutes-level detection efficiency in images of billion-pixel-level size, while achieving high detection accuracy.

Synthetic aperture radar (SAR) images are highly sensitive to observation configurations and exhibit significant variations across different viewing angles, making it challenging to represent and learn their anisotropic features. As a result, deep learning methods often generalize poorly across different view angles. Inspired by the concept of neural radiance field (NeRF), this study combines SAR imaging mechanisms with neural networks to propose a novel NeRF model for SAR image generation. Following the mapping and projection principles, a set of SAR images are modeled implicitly as a function of attenuation coefficients and scattering intensities in the 3-D imaging space through a differentiable rendering equation. SAR-NeRF is then constructed to learn the distribution of attenuation coefficients and scattering intensities of voxels, where the vectorized form of the 3-D voxel SAR rendering equation and the sampling relationship between the 3-D space voxels and the 2-D view ray grids are analytically derived. Through quantitative experiments on various datasets, we thoroughly assess the multiview representation and generalization capabilities of SAR-NeRF. In addition, this article includes few-shot classification performance improvement as a metric for generation performance. The study found that using 12 images per class resulted in an accuracy improvement of nearly 10% for the classification algorithm.

Target recognition for synthetic aperture radar (SAR) images has been a longstanding hotspot. However, using polarimetric information to enhance recognition performance is under-researched. In this paper, we develop a visualization approach to analyze and highlight the contributions of polarimetric elements, and propose a simple polarimetric correlation feature for target recognition. In the visualization method, a channel-wise convolutional structure is well developed, which serves as a proxy of the polarimetric elements. By using the gradients of the target class flowing into each developed convolutional channel with normalization, we obtain activation maps indicating the contribution of each polarimetric element. Then a comprehensive quantitative evaluation of polarimetric element contribution is also performed. We demonstrate again that polarimetric information maintains significant advantages over single-polarization intensity, and the correlations between cross- and co-polarization emerge as key components for target recognition. Exactly inspired by these insights, the straightforward feature is defined as the correlation between the target and typical scatterers, seamlessly fusing intensities and correlations. This polarimetric correlation feature aptly encapsulates the physical scattering of the target, shows a clear mapping relationship with the geometric structure, and captures the differences between categories, achieving intra-category compactness and inter-category separation. Experimental results on both synthetic and real datasets underscore the effectiveness and superiority of this proposed feature.

Target recognition plays a crucial role in the intelligent interpretation of synthetic aperture radar (SAR) images. However, polarimetric information holding great potential in target recognition has not been fully studied. In this paper, we propose a novel method for target recognition in polarimetric SAR (PolSAR) images by using convolutional neural networks (CNNs) to optimize discrete polarimetric correlation pattern. Discrete polarimetric correlation pattern transfers PolSAR images from image domain to rotation domain, and achieves a high-dimensional representation of the target. We then formulate an optimization problem, which is the basis for target recognition, to unfold the low-dimensional embeddings from the raw representations. The optimization problem aims to achieve intra-class compactness and inter-class separation of the target embeddings. Interestingly, we employ CNN as a powerful tool to solve it. By combining the rotation domain features with the neural network, we obtain a discriminative representation that reflects the target’s polarimetric scattering mechanism, and finally realize high-performance target recognition. Experiments performed on both simulated and real datasets demonstrate that the proposed method outperforms reference methods significantly in almost all metrics. It is worth mentioning that even on low-resolution images, the proposed method can still achieve high-precision recognition performance. In addition, through feature visualization, we gain deeper insights into the network behavior. Finally, feature separability issue is also discussed, further confirming that the optimized features do have the characteristics of intra-class compactness and inter-class separation.

Neural network (NN)-based methods have been the mainstream in target recognition. However, the weak interpretability of NNs harbors decision-making risks, which restrict their application in practical scenarios. In this article, we propose to interpret the NN pattern with pruning for target recognition in polarimetric synthetic aperture radar (PolSAR) images. We first generate the initial features of the target by calculating a discrete polarimetric correlation pattern in the rotation domain. In contrast to earlier approaches involving manual extraction of empirical representations, we cast initial feature refining as an optimization problem and employ trainable convolutional layers to address it. Interestingly, the weights of the learnable layers exhibit certain patterns with respect to the rotation angle. To get insights into the NN weights, we use network pruning to highlight the main components of the network weights. In this way, the key polarimetric feature elements can be distinguished, leading to a deeper understanding of the role of polarization information in target recognition. Extensive experiments on Pol-MSTAR and GOTCHA demonstrate that the proposed method not only outperforms existing reference methods in recognition metrics but also greatly provides network interpretability of polarimetric scattering. The correlations between co- and cross-polarization are quite crucial for SAR target recognition.

Amidst the landscape of contemporary remote sensing technology, the endeavor to detect and recognize aircraft within remote sensing images (RSIs) assumes pivotal strategic and practical significance. The complex nature of fine-grained aircraft recognition is a result of the intricate interplay between aircraft and their background environments, alongside category imbalance, which collectively lead to the emergence of a long-tail distribution within the dataset. However, experts proficient in RSIs interpretation can effectively address these challenges through the application of prior knowledge. This paper introduces the Aircraft Reasoning Network (ARNet), a framework tailored for aircraft detection and fine-grained recognition in RSIs, building upon prior knowledge employed in expert interpretation. Specifically, the Knowledge Reasoning Module (KRM) introduces a knowledge graph that incorporates both common and expert knowledge into the end-to-end network. Additionally, the network encompasses a Spatial Context Module (SCM) and an Airport Facility Relationship Module (AFRM). These components facilitate highly accurate detection and recognition of fine-grained aircraft in diverse environmental contexts by employing adaptive prior knowledge reasoning and optimizing target spatial location. Furthermore, an independent Aircraft Component Discrimination Module (ACDM) distinguishes aircraft based on their predominant component features, contributing to improved classification performance in both the few-shot and easily confused categories. Moreover, this paper introduces the AR-RSI dataset, a compilation of RSIs capturing fine-grained aircraft targets from diverse locations. The effectiveness and superiority of ARNet are exemplified on AR-RSI, achieving a minimum of 3.7 percentage higher mAP than the mainstream aircraft detection framework.

This article attempts to infer the orientation angle of the target in synthetic aperture radar (SAR) images using reinforcement learning (RL). It is intended to address the challenges like limited interpretability, the scarcity of SAR data, and complex imaging mechanisms restrict the broader application of learning-based approaches. We propose an interactive deep RL (DRL) framework, where an electromagnetic simulator named differentiable SAR renderer (DSR) is embedded to facilitate the interaction between the agent and the environment. Specifically, DSR generates SAR images at arbitrary orientation angles in real time, helping to simulate a human-like process of angle estimation. The differences in sequential and semantic aspects between images of different orientation angles are leveraged to construct the state space in DRL, which effectively suppress the complex background interference, and enhance the sensitivity to temporal variations. Moreover, to maintain the stability and convergence of our approach, reward mechanisms such as memory difference, smoothing and boundary penalty are incorporated to contribute to the formulation of the comprehensive reward function. Extensive experiments performed on both simulated and real datasets demonstrate the effectiveness and robustness of our proposed method. In addition, when utilized in the cross-domain area, the proposed method mitigates inconsistency between simulated and real domains, outperforming reference methods significantly.

The atomic norm minimization (ANM) method has been well applied in tomographic SAR (TomoSAR) inversion, which can provide accurate scatterer localization and eliminates the outliers effectively. In order to solve the ANM, it is usually converted into a semidefinite programming (SDP) problem. However, this second-order optimization problem suffers from high computational cost when searching for the optimal solution. Since TomoSAR often faces large-scale processing of urban scenes, improving the computational efficiency will benefit greatly. In this paper, we develop and derive an efficient Alternating Direction Method of Multipliers (ADMM) implementation for the ANM method to solve the TomoSAR inversion problem, which is named as the ANM-ADMM algorithm. The detection performance, estimation accuracy, and computational efficiency of the proposed ANM-ADMM algorithm has been carefully analyzed by both simulation and real TerraSAR-X experiments. By comparing with the original ANM-SDP algorithm, it is clear that the ANM-ADMM algorithm can acquire considerable estimation accuracy and can meanwhile improve the computational efficiency significantly.

The simulation of high-resolution synthetic aperture radar (SAR) imagery in intricate environments remains a formidable challenge. Advancements in reversible microwave-domain surface scattering models are crucial, potentially revolutionizing the fidelity of SAR simulations and streamlining the extraction of target parameters. Drawing inspiration from computer graphics, this article proposes a novel differentiable ray tracing (DRT) approach for microwave rendering and fast SAR imaging. The rendering model utilizes coherent spatially varying (SV) bidirectional scattering distribution function (CSVBSDF) based on the Kirchhoff approximation (KA) and the small perturbation method (SPM), corresponding to specular and diffuse scattering contributions, respectively. SAR imaging is efficiently executed via a fusion of ray tracing (RT) and rapid mapping projection. The innovative DRT reversible engine enables swift estimation of SAR image parameter gradients for direct CSVBSDF surface scattering parameter optimization. The method’s validity is confirmed through comparative analysis with measured SAR images and other methods, demonstrating marked improvements in SAR simulation fidelity across diverse observational scenarios by learning surface scattering parameters.

With the development of radar system, there are high demands to perform inverse synthetic aperture radar (ISAR) imaging from sparse frequency bands and sparse aperture data. The compressed sensing (CS)-based and matrix completion (MC)-based methods could be used for analyzing complex targets such as ISAR images of coated scatterers. Compared with the PEC target, complex targets tend to be more noisy and more diverse in image representation, which brings greater initial value offset when calculating the corresponding optimization goals, and thus limiting the performance and effectiveness of the general optimization methods. In this work, we first construct the ISAR coated imaging model from the aspects of PEC scatterer, then a fast low-rank matrix completion and norm regularization (MC+NR) algorithm is proposed for high-resolution coated ISAR sparse imaging, which can effectively preserve target features while maintaining sparsity. The experiments using simulated data demonstrate that the proposed method converges over 5 times faster than norm regularization method and can obtain a better sparse imaging result as well.

The article proposes a generalized optimal sensing principle for synthetic aperture radar (SAR) imaging, maximizing the mutual information between the sensed object and the reconstructed image with the optimal SAR measurement matrix. Inspired by Shannon’s capacity theorem, the SAR sensing capacity is derived, which represents the maximum mutual information that can be acquired per unit distance or unit area in 1-D or 2-D scenarios. The SAR sensing capacity serves as a theoretical performance bound, guiding the design of SAR sensing systems and enabling reasonable estimation of systems’ performance. Additionally, this article analyzes the relationship between system parameters and the column correlation of SAR measurement matrices, guiding the design of the SAR system. Furthermore, the optimal sensing principle is applied to the variable-resolution SAR (VR-SAR) imaging system. Theoretical simulations are conducted to verify the feasibility of the optimal sensing principle, and examine the relationship between column correlation and SAR parameters. Additionally, the advantages of VR-SAR based on the optimal sensing principle are compared with those of the conventional strip-map SAR mode.

Synthetic aperture radar (SAR) has become a state-of-the-art technology in many applications without being affected by changes in weather and daylight. Since the detection capability of the single-dimensional SAR is limited, the multidimensional (MD) SAR system, e.g., multibaseline and multipolarization, is used to improve its performance. The design of the MDSAR system should be directly related to the specified applications and a quantitatively analytical theory for bound analysis is required to achieve good efficiency. In this article, a mathematical framework for inversion bounds analysis of MDSAR is proposed. First, based on the attributed scattering center (ASC) model, the Fisher information matrix and its corresponding Cramer–Rao lower bound (CRLB) are used to get the error bound of the estimated parameters. Second, considering the discrete sampling (DS) of the parameters, a probability density function-based conversion is conducted to get the DS CRLB. Finally, the mathematical framework for MD acquisitions is established. The simulation-based experimental results show that the theoretical error bound is consistent with the output of the orthogonal matching pursuit (OMP). The error bound of the parameters obtained by the proposed general mathematical framework can be used to evaluate the performance of inversion algorithms under certain MDSAR configurations.

Dust particles in Martian atmosphere can significantly reduce visibility. This article proposes a physically guided neural network approach for dust removal of Martian images by incorporating atmospheric scattering model into the cycle-consistent generative adversarial network (Cycle-GAN) framework. The network consists of two primary modules: dust removal and dust addition, both of which combine neural networks and physical models. The dust-removal process estimates the scattering coefficient and transmission map and then physically implements dust removal with the atmospheric scattering model. The dust-addition process estimates the scene depth, which is then combined with the atmospheric scattering coefficients obtained by the dust-removal process to compute the transmission map. The transmission map is substituted into the atmospheric scattering model for dust addition of clean images. Moreover, the additional loss of transmission map and scattering coefficient furtherly enhances the consistency constraints. Martian clear and dust images collected by the Mars Curiosity rover are used to train and evaluate the new dust-removal approach. Extensive ablation experiments demonstrate the effectiveness of incorporating the physical model and the additional loss functions. Furthermore, the scattering coefficients learned by the network are validated with the Mie scattering theory, ensuring the physical plausibility of the estimated parameters. The key advantage of this approach is that it does not require paired images of the same scene with or without dust, which is often a limiting factor for supervised dust-removal techniques. By incorporating the physical scattering model into the Cycle-GAN framework, the network can learn to restore clear images from dust-affected ones in a more realistic and consistent manner.

2D SAR image suffers from distortions such as layover, foreshortening and shadow. These distortions can be well resolved if the surface elevation is determined. However, traditional single-pass InSAR is limited by the phase continuity assumption, which can be mitigated by using the multi-baseline observations. Fortunately, airborne array-InSAR system acquires the multi-baseline in a single flight, but side-lobe will significantly affect the performance of 3D reconstruction. Sparse radar imaging is an efficient way to suppress the sidelobe. In this paper, a capon-based approach is used to reduce the sidelobe and improve the quality of 3D reconstruction. In the practical application, the performance of sidelobe reduction and weak signal protection is contradictory. A trade-off is made by using the similar indicators to get the optimal subaperture size. Experimental results by the real airborne array-InSAR images show that the reducing sidelobe improve the coherence of the weak scatterers, improving the performance of 3D reconstruction.

Lack of interpretability and weak generalization ability have become the major problems with data-driven intelligent Synthetic Aperture Radar-Automatic Target Recognition (SAR-ATR) technology, especially with sparse training samples. A novel insight into the typical target representation with neural networks from a causal perspective is presented in this paper. Specifically, a set of SAR images is causally modeled by intrinsic, diverse, and random attributes, which is consistent with the forward imaging process of SAR. Consequently, an Interpretable Disentangled Adversarial auto-Encoder (IDAE) is proposed based on the disentangled representations. A Symmetrically Conditional Encoding (SCE) module is established to constrain the semantic consistency of low-dimensional features. Besides, a hybrid loss function is designed for iterative training. Experiments conducted on the MSTAR dataset show that the proposed model improves both representation and generalization abilities. IDAE is able to achieve a classification accuracy of 93.1% using only 12 samples per class.

Current synthetic aperture radar (SAR) missions with the satellite constellation can effectively reduce repeat time, achieving a near real-time deformation monitoring. In the practical application, dynamic InSAR process is a good option to improve the computational efficiency. The proposed amplitude-augmented recursive InSAR time series enables both deformation anomalies and surface changes detection, which gives a new route to analyze the InSAR data. And quality indicators such as detect-ability power and the minimum detectable deformation are proposed to interpret the detected anomalies. However, the the sensitivity of the anomaly detection depends on the choice of the processing strategy. A quantitative analysis of the sensitivity is required. In this work, the key part of the temporal deformation anomaly detection, such as the number of update observation and quality metrics is investigated, which gives the users a reasonable demonstrations in the practical applications.

Synthetic Aperture Radar (SAR) is extensively employed in both civilian and military sectors, with recent advancements leveraging deep learning for automatic SAR image interpretation. However, the effectiveness of these techniques, particularly Convolutional Neural Networks (CNN), is challenged by insufficient angle range in actual samples due to satellite incident angle constraints. This article proposes a method for generating multi-view samples of SAR targets based on a CNN module integrated with Differentiable SAR Renderer (DSR). Specifically, a polygon mesh is reconstructed from two-dimensional (2D) SAR images through the CNN module, and the DSR is utilized to reversely render SAR target images of various viewpoints from the reconstructed mesh, including the samples used to match with original input 2D images. Then, the generated images is used to compute the loss in training phase, and no three-dimensional (3D) ground truth is required. Experiments are conducted on simulated SAR images and the results demonstrate the efficacy of multi-view sample generation for SAR targets.

The model performance of object detection task may dramatically deteriorate when meeting the new dataset with discrepant data distribution compared with trained images. Especially for Synthetic Aperture Radar (SAR) images, the complicated imaging mechanism and diverse environments probably induce intense changes in image appearance and hurt the detection capability and robustness of models based on deep learning. In this paper, a method of learning strong interference characteristics of SAR images is proposed and conducted to generate artificial SAR images as extra training samples in the downstream task -- object detection to improve the detection accuracy and decrease the missing rate of models. Our approach utilized as a data augmentation strategy without annotation cost is confirmed to be efficacious and reliable by multiple experiments.

In real-world recognition of remote sensing (RS) targets, large amount of RS data is hard to be acquired at once, but arrives in batches, which means the constantly adaption of models for new data and new classes. However, training old and new data together from scratch has certain requirements on data storage space and retraining time, so incremental learning come to be desirable for future RS recognition systems. In this article, a class incremental learning method based on knowledge distillation is proposed for RS image classification. In order to better retain the knowledge of old tasks, a relation-based loss is used as a new matching manner in distillation loss, which frees the student model from the burden of matching the exact output of the teacher model. Experiment results based on UC Merced 21, NWPU-RESISC45 and plane objects of FAIR1M demonstrate the advantages of the proposed method.

Simulation of Synthetic Aperture Radar (SAR) image in complex scenes has always been a challenging research. The key step of the simulation is determining the parameters of the bidirectional scattering distribution function (BSDF). However,the lack of BSDF parameters optimization make it difficult to get a good simulation. In this work, an optimal parameter estimation of BSDF based a differentiable ray-tracing is proposed. In this SAR image simulation engine, ray-tracing mapping and projection algorithm (MPA) can be inversely differentiable and thus gradient estimation of parameters can be quickly obtained from simulated SAR images. Additionally, with the help of robust physical scattering model based on small perturbation method (SPM) for microray tracing, a better BSDF is obtained, which improve the performance of SAR simulation.

Above Ground Biomass (AGB) estimation is a basis for rational utilization of natural resources and ecological succession process. Recently, multiple sources of remote sensing data have been used to estimate AGB at high spatial resolution, overcoming the limitations of each type of data. In order to fully exploit the potential of deep learning models based on multi-source data in AGB estimation, an end-to-end Deep Neural Networks (DNN) model is developed using Sentinel-1/2 data, and a learnable weight matrix is designed to tune the contribution of different predictors in multi-source data, thus improving the performance of the model. The experimental results show that the designed DNN model can achieve accurate AGB estimation with a coefficient of determination of 0.7314. Compared with the widely employed XGBoost and Random Forest machine learning models, the proposed DNN model has improved by 6.29% and 5.07% for grassland biomass estimation, respectively.

Accurate prediction of Tropical cyclone (TC) intensity using multispectral images (MSIs) is critical to avoid economic loss and life casualty. Although existing methods have achieved good prediction results, they neglect changes in cloud patterns such as cyclone eyes and cloud spirals, which are closely related to TC intensity. How to leverage temporal-spatial-spectral features of MSIs to improve prediction accuracy is challenging task. In this paper, we propose a novel framework with Spectral-Temporal Dislocation and Attention-Based Networks (STD-AN) to predict MSW speed values near cyclone centers. The STD technique allows the framework to learn temporal-spatial-spectral features of TC. Meanwhile, the Self-Attention Modules (SAM) enable global attention feature extraction and Cross-Attention Modules (CAM) fuse different band features to improve prediction accuracy. Experimental results show that the proposed framework outperforms several state-of-the-art methods for TC intensity prediction.

The interferometric vertical wavenumberkzhas a nonnegligible impact on the measurement accuracy. A critical study on the optimumkzfor forest height mapping must be carried out. This paper quantitatively investigates the PolInSAR inversion performance by the Cramér-Rao Lower Bound analysis. Through studying the relationship between the volume coherence and the inversion performance, a volume coherence condition for the existence of the optimumkzis first proposed for L-band PolInSAR inversion. The theoretical optimumkzcan then be easily obtained from the constant volume coherence level. We demonstrate that the established volume coherence condition can be useful for the system designers to optimize the system configurations before the PolInSAR mission.

Three-dimensional (3D) reconstruction of complex targets on the ground from multi-view synthetic aperture radar (SAR) images is of great interests. The inherently-integrated forward-inverse architecture of the differentiable SAR renderer (DSR) provides a promising solution to the general inverse problem of SAR target reconstruction. In this context, the target’s shadow provides complementary information to its scattering image. Hence, this paper proposes a novel DSR-based target reconstruction approach using both the target image and its shadows. The capabilities of DSR are extended to generate not only target scattering images but also shadows. Furthermore, the gradients of the outputs, specifically illumination map and shadow map, with respect to the inputs, i.e., target geometry represented as a mesh, are derived. This enables us to develop a gradient-descent inverse approach for solving the general reconstruction problem. Extensive simulations and quantitative evaluations demonstrate that incorporating both the target scattering image and its shadows significantly improves the reconstruction performance. Moreover, our analyses indicate that achieving optimal reconstruction effects requires a minimum of 9 views with a relatively even distribution. Finally, the proposed algorithm is validated using real SAR images of vehicle targets.

Lack of interpretability and weak generalization ability have become the major challenges with data-driven intelligent SAR-ATR technology, especially in practical applications with azimuth-sparse training samples. A novel insight into SAR image representation with neural networks from a causal perspective is presented in this paper. Firstly, a causal model of SAR image representation conditioned on disentangled semantic factors is proposed. A set of SAR images is considered as a low-dimensional manifold, which is controlled by three semantic factors, namely, intrinsics, diversity, and randomness. A Causal Adversarial auto-Encoder (CAE) for SAR-ATR is then proposed to embody this disentangled representation, which incorporates a number of novel built-in network features. A physically reasonable Cyclic High-frequency information-based Embedding (CHE) method is proposed for azimuth encoding, which ensures the uniformity, continuity, periodicity, and distinctiveness of angle. A Symmetrically Conditional Encoding (SCE) module is established to constrain the semantic consistency of low-dimensional features. Besides, a hybrid loss function is designed, which is composed of latent adversarial loss, reconstruction loss, and task-oriented losses. Both representation and generalization abilities are thoroughly evaluated through qualitative visualization and quantitative comparison experiments on the MSTAR and FUSAR-Ship datasets. Experimental results demonstrate superior representation ability for the disentangled properties via angle-interpolation and target-transformation of SAR images. By using only 12 samples per-class, the proposed CAE can achieve an accuracy of 93.1% for the 10-target SAR-ATR classification task.

Speckle reduction is a longstanding topic for polarimetric synthetic aperture radar (PolSAR) images. In this paper, we propose a novel end-to-end PolSAR image despeckling framework for the first time, which predicts the weight matrices of neighboring pixels instead of the target pixel itself nor the nor the noise, to achieve image despeckling. It hardly relies on any assumptions on the speckle noise distribution. Within this framework, residual in residual scaling network (RIRSN) is developed by combining the advantages of residual connections and residual scaling. To reduce network redundancy further, a dynamic version of RIRSN (DRIRSN) is also proposed by adjusting the network structure dynamically based on noise level and image content. Specifically, in DRIRSN, we introduce a lightweight network called picture2vector to estimate noise level, and a well-designed loss function to estimate image information level and measure image denoising quality simultaneously. The proposed picture2vector and loss function guide DRIRSN to focus on image areas with rich content and information, enhancing the adaptability of the network. DRIRSN inherits the properties of RIRSN for adaptively selecting and weighting the pixels of the neighborhood, and dynamically adjusts the network structure according to the estimated noise level and image content. We compare the proposed networks with reference methods on both simulated images and real images. Experimental results demonstrate that the proposed networks can effectively reduce speckle noise with low time consumption and, meanwhile, better preserve the details and the repetitive structures such as textures and edges, and the polarimetric scattering characteristics, compared with the other methods.

This article takes on the challenges of synthetic aperture radar (SAR) imaging for miniaturized SAR (MiniSAR) onboard a multirotor unmanned aerial vehicle (UAV). Several unique challenges are systematically analyzed, and a corresponding analytical phase error model is established, which accurately models the effects of both translational and rotational motions of UAVs. A segmental aperture imaging (SAI) algorithm, an autofocus algorithm based on strong scatterers, is proposed. It simply divides the platform trajectory into uneven segments, which are first independently focused with motion compensation and then stitched together to form a complete SAR image. Both the theoretical derivation of the signal model and the implementation of the imaging algorithm are presented. A simulation analysis with actual UAV trajectory and attitude data is conducted, which demonstrates the efficacy and performance of the proposed imaging algorithm. It shows that the ideal focusing effect can be achieved as evaluated by various metrics, and the proposed algorithm has superior performance compared to the subaperture phase gradient autofocus (PGA) and minimum entropy autofocus (MEA) methods. Finally, the multirotor-borne MiniSAR system FUSAR-Ku is used for experiments to verify the proposed algorithm. Experimental results show that the proposed algorithm can achieve the theoretical decimeter-resolution imaging performance as measured by various metrics.

Convolutional neural networks (CNNs) are widely used in image classification, but such methods often require massive labeled data as learning resources. On the one hand, synthetic aperture radar (SAR) image interpretation is difficult, resulting in the lack of large-scale data sets with high-quality labels. On the other hand, CNNs are not explainable enough to provide reliable and trusted application services for SAR target recognition. To solve the above problems, physics-based electromagnetic simulated images are used to alleviate the shortage of real data with annotations, and explainability analysis methods are introduced to explain the basis of network decision-making. To address the domain gap between simulated and measured data, we propose a novel network integrating dual-branch image reconstruction and subdomain alignment (DBRSA). The network completes the reconstruction of simulated and measured images through the domain-shared encoder and domain-specific decoders, thereby helping the encoder to learn feature extraction methods independent of labels. In addition, the network aligns the feature vectors of similar targets obtained from different domains according to the real or pseudo labels of the samples, so as to further improve the classification accuracy. The experimental results and model decision analysis results demonstrate that the proposed network can improve the performance reliably by reducing the attention to the background noise and increasing the attention to the shadows and contours of the target, effectively reducing the dependence on the number of sample labels in practical application scenarios.

Synthetic Aperture Radar (SAR) images are impaired by the presence of speckle. Despite the deep interest of scholars in the last decades, SAR image despeckling is still an open issue. Among different approaches, recently, many Deep Learning (DL) methods have been proposed following both supervised and unsupervised training approaches. There are two main challenges within the supervised framework: training data, and cost functions. Our approach builds training datasets which are varied and realistic using a multi-category Generalized Gaussian Coherent SAR simulator. It allows modeling a variety of SAR scenarios beyond the fully developed speckle hypothesis, which is only valid in homogeneous areas. Such multi-category simulated speckle is then applied to a noise-free reference obtained by multi-looking a temporal stack of actual SAR images in order to obtain the noisy input. We design an effective multi-objective cost function that accounts for texture, edge, and statistical properties preservation. We show the superiority of our approach assessing numerically and quantitatively its performance with three different SAR datasets.

To address the challenges in synthetic aperture radar (SAR) ship target detection, this paper proposes a SAR ship small target orientation detector named FADet based on semantic flow feature alignment and Gaussian label matching. First, to solve the feature misalignment problem caused by feature extraction downsampling and residual connections, we introduce the FAM module into FPN, which automatically aligns deep and shallow fine-grained semantics information through semantic flow alignment. Second, due to the scattering characteristics of SAR imaging, the boundary information of SAR targets is not obvious, we combining attention mechanisms design an adaptive boundary enhancement module to enhance the target boundary information. Finally, to solve the problem that small targets have difficulty matching positive samples under IOU rules, we design a label matching strategy based on Gaussian distribution. This matching strategy can still learn regression information when two boxes do not intersect. Based on the SSDD+ and RSDD-SAR datasets, the effectiveness of each module in FADet is verified by ablation experiments. Additionally, through comparison experiments with the latest orientation detection methods, FADet achieves a good compromise between accuracy and inference speed. The AP50 and AP75 on the SSDD+ and RSDD-SAR is 91.03, 59.94 and 90.78, 59.91 respectively, and the FPS is 19.83.

Quadric surfaces are commonly seen geometries in man-made targets. In this paper, a unified bidirectional scattering distribution function (BSDF) is analytically derived for general convex quadric surfaces including both the doubly- and the singly-curved surfaces. Based on physical optics (PO) and stationary phase method (SPM), the BSDFs of the doubly- and singly-curved surfaces are first deduced separately. Then the unified form of the two BSDFs is formulated, which can smoothly degenerate to any specific type of canonical curved surfaces by taking the corresponding values of the geometric parameters. Comparison with numerical PO demonstrates the correctness and efficacy of the proposed model. This model can be used to continuously model the bistatic polarimetric scattering behavior of a localized quadric surface patch. It can be used as the prototype for a scattering dictionary for both forward and inverse problems of electromagnetic scattering, which is of great value to radar target recognition and radar image interpretation.

Array-InSAR system can acquire the multi-baseline images in a single flight, which significantly improves the capability of the practical application. However, the array-InSAR system with many channels has high complexity in both system and data processing due to the cross-channel calibration. Investigation of the 3D reconstruction suitable for sparse array-InSAR images enables the reduction of the number of channels. This work propose evaluates the possibility of the 3D reconstruction using three channel array-InSAR images. To improve the reliability, the proposed 3D phase unwrapping (PU) works on a framework combining short and long baseline interferograms. Additionally, Using a success unwrapping criteria, the bounds of th possible baseline combination is derived. Experimental results by real SAR data show that the proposed method is able to achieve 3D reconstruction in urban areas with high precision using only three channel array-InSAR images and optimize the baseline design of the array-InSAR system.

SAR target detection in high-cluttered environments is a challenging task, especially in medium-resolution SAR images. It is hard to identify targets due to strong background inferences. However, targets often change over time while most background scatterers are temporally stable. With the increasing availability of multi-temporal SAR images, one can separate moving targets from the background by analyzing their temporal behavior. In this paper, a novel method is proposed to remove background clutters by developing a stable background mask. First, we investigate the temporal behavior of amplitude time series (TS) of different scatterers using three typical distributions. Then, temporally stable pixels are extracted to generate a stable background mask for the scene. We block the stable regions for each image in the amplitude TS and obtain the processed images where only unstable regions that may include targets are left. Since most background inferences are eliminated, it will help lower false alarms in the target detection procedure.

The circular polarization ratio (CPR) was defined in compact-polarization (pol) mode as an indicator of water-ice in lunar PSR (permanently shadowed region). CPR is a composite pol-parameter described by co-pol and cross-pol scattering components, which caused by surface roughness and rocky objects on surface. In this paper, CPR is derived with linear-pol and circular-pol scattering components. Radar echoes from different rough surfaces are numerically simulated with the bidirectional analytic ray tracing (BART) method. The CPR, degree of polarization$(m)$and the relative phase$(\delta)$are numerically presented. As an example, Mini-RF radar images of the PSR crater Hermite-Band no- PSR crater Byrgius C are analyzed to illustrate how the roughness lead to different CPRs inside and outside the intermediately degraded craters.

The interpretability and generalizability of neural networks are well-aware issues in traditional deep learning due to the black-box nature of pure data-driven neural networks. While the physics-inspired neural networks (PINNs) can achieve a more generalized supervised model under few-shot learning by taking the physical principles as prior information into the network design. Especially in the electromagnetic field, the applications and realization of the PINN based on electromagnetic information are important. It can help solve the few-shot learning problems and improve the generalization of deep learning for electromagnetic data. This paper introduces several PINN models for electromagnetic problems, which can significantly reduce the reliance on the training sample size under the same level of networks parameters.

This paper addresses the unique motion errors that must be compensated for miniaturized synthetic aperture radar (MiniSAR) onboard multi-rotor unmanned aerial vehicle (UAV). Segmental aperture imaging (SAI) algorithm is proposed which simply divides the platform trajectory into segments. Each segment is first independently focused with motion compensation and then stitched together to form a complete SAR image. A real-time SAI algorithm implementation is proposed. It performs overlapping subaperture for each segment and performs polynomial fitting of its phase. It is implemented on graphics processing unit (GPU). Finally, the data obtained by the FuSaR-Ku system is used to verify the effectiveness of the real-time SAI algorithm.

With the development of$5\mathrm{G}$, the location of base stations is related to the electromagnetic wave scattering with the surface of objects, especially the building walls in cities. In this study, the building wall is simply modeled as a two-dimensional plane surface. Since the roughness of the building wall is relatively small, the specular reflection is the main component of the scattering wave. The bidirectional reflectance distribution function (BRDF) of building walls is approximated as the Fresnel reflection coefficients. Then, a field measurement system, called FUSAR-Rail-S, is established to measure the scattering of the building wall. In the measurement, two different walls are used to analyze the roughness effect on the scattering coefficient. A comparison of the measurement result with the Fresnel reflection coefficients shows that they are in good agreement in most angles. Within a certain angle range, the reflection coefficients increase when the incident angle becomes larger. In addition, the reflection coefficients of the smaller roughness wall are larger.

With the continuous development of Deep Learning (DL), ship detection in SAR (Synthetic Aperture Radar) images based on convolutional neural networks (CNN) has become a common approach. CNNs with complex structures have achieved good performance in SAR images, but face challenges such as high time consumption and high false alarm rate, because of the sparsity of ships in the remote sensing images. In this paper, a rotated ship detection network based on the CFAR (Constant False Alarm Rate) preprocessing is proposed to address these problems. It first uses a CFAR preprocessing to fast narrow down the scope of detection so as to save processing time. Then, a classification network is designed to reduce the false alarm rate. The experiment results based on the Gaofen-3(GF-3) dataset show that the proposed method can reduce the false alarm rate greatly and use much less CPU time.

Compared with visual sensors that have light dependence and privacy intrusion issues, non-intrusion millimeter-wave (mmW) radars are more suitable for the daily person identification. In a realistic home scenario, there are new challenges that are not taken into account in the existing research. This letter attempts to address these issues such as multipath interference, complex walking process, and recognition robustness in smart home scenarios and designs a lightweight multi-branch convolutional neural network (CNN) with an Inception-Pool module and a Residual-Pool module to learn and classify gait Doppler features. The experimental results in a home living room scenario indicate that the designed mmW radar person identification system can achieve accurate and robust real-time identification performance.

Data augmentation is beneficial when the measured training data are insufficient to train a robust deep model. One of the promising techniques is to use simulated data generated by physics-based engines. For example, few-shot learning of synthetic aperture radar (SAR) targets could be benefited from simulated SAR images. However, the characteristics of the simulated training data significantly affect the performance of the trained model. Therefore, it is of great significance to evaluate the utility of simulated data objectively and effectively. A recognition rate versus substitution rate curve (RSC)-based assessment criterion is proposed, consisting of substitution rate (SR)-based dataset allocation stage and RSC-based evaluation stage. First, the differential dataset allocation is performed under a progressive SR to obtain paired reference and comparison training sets. Then, the reference and comparison classifiers are trained under different SRs using the same network and parameter configuration in the RSC criterion-based evaluation stage. AconvNet and AlexNet are selected as the backbones of the evaluation network. Especially, k-fold cross-validation is applied to alleviate selection bias. The difference between the integrals of RSCs is defined as the RSC score for the simulated dataset. Experiments conducted on the measured and simulated moving and stationary target acquisition and recognition (MSTAR) database demonstrate the rationality and validity of the proposed RSC criterion. Specifically, multisource simulated datasets are adopted, including the adversarial autoencoder-generated and electromagnetic simulation datasets. The proposed RSC criterion shows promising utility evaluation ability, flexibility, and extensibility compared with traditional full-reference image-quality assessment criteria.

Multi-temporal synthetic aperture radar interferometry (MT-InSAR) is able to reconstruct a 3D surface model with high precision but requires a long waiting time to get the multi-baseline SAR images. Array-InSAR system can acquire multi-baseline images in a single flight, which significantly improves the practical capability of 3D reconstruction. However, the array-InSAR system with many channels has very high complexity in both system and processing because of cross-channel calibration and decoupling. Thus, reducing the number of channels requires the investigation of the 3D reconstruction algorithm to be suitable for sparse array-InSAR images. This work proposed an asymptotic 3D phase unwrapping (PU) algorithm for 3D reconstruction using sparse array-InSAR images, i.e., as few as three or four channels. A 2D (space) + 1D (baseline) PU framework is developed to improve the reliability of the 3D PU and a novel asymptotic strategy is proposed for the combination of the short-long baseline interferogram. Using a successful unwrapping (SU) criteria, the bounds of the possible baseline combination and the expected minimal coherence are derived, respectively. The main advantage of the proposed algorithm is the reliable phase unwrapping with very sparse channels and an analysis of the possible baseline combination. The experimental results by both simulated and real data show that the proposed method can achieve a 3D reconstruction using only three-pass array-InSAR images and optimize the baseline design for the array-InSAR system.

Microwave imaging with large rotating angle and sparse sampling is an attractive approach to obtain the high-resolution target image with reduced radar resource. However, the popular imaging methods, e.g., Range-Doppler (RD), back projection (BP), and sparse recovery (SR), are difficult to deal with large rotating angle and sparse sampling simultaneously. In recent years, deep learning (DL) has been widely studied and been successfully used to handle the problems in computer vision. However, since most existing DL networks are put forward for the real visual image and a large amount of data is essential for network training, DL cannot be directly used to process the complex and sparse target echo for microwave imaging. In this article, a new learning imaging framework is proposed and a model-driven learning imaging network (MDLI-Net) is built for high-resolution microwave imaging with large rotating angle and sparse sampling. In the proposed framework, the electromagnetic scattering model is used to generate the training data efficiently, and the sparse microwave imaging theory is applied to guide the design of the deep imaging network. By inputting the 2-D sparse complex-valued target echo, the trained MDLI-Net can output the high-resolution and focused target image efficiently. The effectiveness of the proposed learning imaging method is validated by experiment results with both simulated and real data.

Multitemporal synthetic aperture radar interferometry (MT-InSAR), capable of detecting both surface deformation and elevation with high precision, is used for many applications in earth observation. Conventional synthetic aperture radar (SAR) missions with a single beam only detect deformation along the line of sight (LOS) and relative elevation due to the undetermined model of phase wrapping. In a multisatellite SAR mission, measurements from different SAR geometry improve the sensitivity of the detectable deformation, especially to the deformation along the north–south (N-S) direction. However, it is difficult to combine the measurements from varying viewing angles since the absolute phase cannot be reconstructed without a ground control point. In this article, a tri-beam SAR system is introduced to detect 3-D deformation and derive multiview 3-D surface model from a single spaceborne platform. The accuracy of the 3-D deformation from the tri-beam SAR is exploited with varying squint and incident angles to obtain the optimal parameters of the three beams. Then a multidimensional coherent scattering model is used to simulate the multitemporal SAR data with different viewing angles. Regarding the tri-beam SAR, potential applications in earth observation including 3-D deformation monitoring, geodetic stereo SAR, and multiview 3-D forest reconstruction are investigated subsequently. The results of this study indicate that the tri-beam SAR is able to measure 3-D deformation and reconstruct 3-D surface model without ground control point.

In airborne array interferometric SAR (Array-InSAR) tomography, the measurements acquired by conventional uniform sampling array are always restricted by the number of physical baseline elements and the size of baseline aperture. It is desirable to capture new acquisitions and enlarge the aperture with virtual signal processing instead of actually adding array baselines. For this motivation, we utilize the disparity of a pair of coprime sampling sub-arrays to enlarge the baseline aperture and construct new observations virtually. The generation of virtual measurements is equal to estimating cross-correlation matrices in real SAR data. Due to the spatial target variation, we adopted an adaptive filtering method to estimate the cross-correlation matrix. We call the above-mentioned processing of generating virtual measurements as acoprime sensing technique. The newly generated virtual measurements have more degrees of freedom, a larger baseline aperture, and a higher signal-to-noise ratio (SNR) than the physical measurements. These advantages offer the possibility to obtain competitive three-dimensional (3-D) imaging results without increasing the hardware cost of the Array-InSAR. We demonstrate the effectiveness of the proposed method by the coprime acquisitions selected from AIRCAS Array-InSAR data.

Deep learning-based synthetic aperture radar (SAR) target recognition often suffers from sparsely distributed training samples and rapid angular variations due to scattering scintillation. Thus, data-driven SAR target recognition is considered a typical few-shot learning (FSL) task. This article first reviews the key issues of FSL and provides a definition of the FSL task. A novel adversarial autoencoder (AAE) is then proposed as an SAR representation and generation network. It consists of a generator network that decodes target knowledge to SAR images and an adversarial discriminator network that not only learns to discriminate “fake” generated images from real ones but also encodes the input SAR image back to target knowledge. The discriminator employs progressively expanding convolution layers and a corresponding layer-by-layer training strategy. It uses two cyclic loss functions to enforce consistency between the inputs and outputs. Moreover, rotated cropping is introduced as a mechanism to address the challenge of representing the target orientation. The moving and stationary Target recognition (MSTAR) 7-target dataset is used to evaluate the AAE’s performance, and the results demonstrate its ability to generate SAR images with aspect angular diversity. Using only 90 training samples with at least 25° of orientation interval, the trained AAE is able to generate the remaining 1748 samples of other orientation angles with an unprecedented level of fidelity. Thus, it can be used for data augmentation in SAR target recognition FSL tasks. Our experimental results show that the AAE could boost the test accuracy by 5.77%.

3-D computerized ionospheric tomography (CIT) is an ill-posed problem due to the insufficient amount of observations, it remains challenging for practical applications. In this article, we proposed an ionospheric tomography method that combined data-driven methods with compressed sensing (CS) to deal with the ill-posed problem. First, slant total electron content (STEC) data were extracted by undifferenced and uncombined precise point positioning (UCPPP) with known fixed station coordinates. Second, data-driven methods were adopted to construct the projection matrix from the ionospheric model. Third, compressed sensing was used to derive the sparse solution based on$L_{1}$norm. The ionospheric tomography can be achieved well by using observations during the shorter time interval and in a sparse receiver distribution based on the property of compressed sensing. Results of experiment based on real Global Positioning System (GPS) observation data verified the effectiveness of the proposed methods. By comparing with the colocated ionosonde, it is found that the CS methods are more consistent with the actual ionospheric fluctuation than the modified constrained algebraic reconstruction technique (CART). In terms of the differential STEC (dSTEC) analysis, the error of the tomography model by Compressed Sensing-Principal Component Analysis (CS-PCA) is less than 0.2 TEC unit (TECU), and the time resolution is 5 min. The UCPPP with constraint by CS-PCA shows the best performance of 12.2%, 40.9% and 0.31% improvement in positioning accuracy, convergence time, and fixed rate over the UCPPP with constraint by modified CART. The proposed data-driven methods may be important for high-resolution 4-D ionospheric tomography in the future.

Synthetic aperture radar (SAR) tomography (TomoSAR) extends the synthetic aperture principle into the elevation direction for 3-D imaging. Due to the sparsity of the elevation signal, the compressive sensing (CS) methods have been introduced for tomographic reconstruction. However, the limited irregular acquisitions and the dense sampling grids of the elevation cannot guarantee the sufficiently sparse reconstruction in the presence of noise. By constructing a complete set of atoms, the gridless sparse methods can directly recover the sparse signals in the continuous frequency space. In this paper, we propose the Atomic-norm minimization or the Gridless CS approach for tomographic SAR inversion and compare it with the L1-norm based optimization. The enhanced sparsity, the super-resolution capability and the more accurate estimates are demonstrated using the numerical simulations and experiments with real data. A Gridless CS reconstruction of an urban area of Shanghai from the TerraSAR-X data set are presented.

When detecting the underground structure of Mars using an orbital subsurface investigation radar, it is essential to compensate for the ionosphere distortion of the received echo. The existing algorithms, such as the contrast method (CM) and the phase-gradient-autofocus (PGA) algorithm, can mainly compensate for the defocus of the echo, but it is difficult to compensate for the group delay at the same time, resulting in the deviation between the echo position and the terrain profile. This article analyzes the interference of Martian ionosphere to the echo of subsurface investigation radar and proposes a dual-band difference (DBD) ionosphere compensation algorithm. First, the group delay model of radar echo under ionospheric interference is established. Then, the equivalent plasma frequency of the ionosphere is calculated according to the group delay difference of echoes in two bands, and the Taylor coefficients of the phase error are calculated. Finally, the echoes are accurately compensated with the Taylor series model of phase shift. This algorithm can correct the defocus and sidelobe interference and simultaneously compensate for the group delay, which is greatly significant to accurately locate the subsurface structure of Mars. The experiments on the simulation data and the echo of the Mars Advanced Radar for Subsurface and Ionosphere Sounding (MARSIS) show the feasibility of this method. Compared with the traditional way, this algorithm is suitable for both high- and low-frequency situations, and the high computational efficiency without iteration makes on-orbit real-time ionosphere compensation possible. Moreover, with the effective correction of topographic offset, the total electron content (TEC) can also be accurately retrieved.

A generalized synthetic aperture radar (SAR) modality of operation named variable-resolution (VR) SAR is proposed, which explores the diversity of antenna patterns and inhomogeneity of pulse repetition frequencies (PRFs) for adaptive imaging. Based on the relationship between the azimuth resolution and the corresponding integration angle, it uses dynamic beam patterns along the trajectory to illuminate different regions. We formulate the optimization problem of VR SAR based on the principle of maximum mutual information. First, the information content of a specific scene is defined by modeling its distributed scattering as a stochastic process, and the mutual information between scenes and the observed SAR image can be derived. Then, we construct an optimization problem to maximize the mutual information by solving for the optimal beam manipulation scheme of the VR SAR mode. Further optimization of PRF is conducted to obtain a relatively larger swath width and smaller data volume by compromising the resolution of some regions with less information while ensuring there are no azimuth ambiguities. The potential advantages of VR mode are: 1) it simultaneously provides higher resolution for the high-information regions and a larger imaging area than would otherwise be possible in strip map and spotlight SAR modes and 2) it optimizes the efficiency of data acquisition while extracts as much information from scenes as possible. A mathematical model of VR SAR mode with the principle of maximum mutual information is established, and the feasibility and merits of the method are demonstrated through a series of simulations and an equivalent experiment using RADARSAT-1 raw data.

This paper establishes a low degrees-of-freedom allometric vegetation model based on the Biomass and allometry database (BAAD). It consists of a 5-bit species encoding scheme and a 5-parameter relationship derived from the BAAD data records. Combined with opensource fractal tree generation engine, it can produce realistic tree samples of a large variety of species. A coherent electromagnetic scattering calculation method is developed for the vegetation model where the generalized Rayleigh-Gans (GRG) approximation and the infinite cylinder approximation are used to calculate the scattering matrix of leaves and branches/trunk. Four-path multiple scattering mechanisms between vegetation and the ground are considered and attenuations through vegetation canopy are also considered. The scattering model is validated against numerical methods. In addition, an end-to-end simulation tool is developed. Optical image is used to extract individual trees with center positions and crown diameters. The rest parameters are generated using the derived allometry model. Virtual 3D scene with vegetation on digital elevation map (DEM) can be generated and SAR images can be simulated. Several case studies are carried out for both polarimetric synthetic aperture radar (SAR) interferometry (PolInSAR) and multi-temporal interferometric SAR (InSAR). One case of the Mount Fuji area is simulated and validated against ALOS-2 data and demonstrates an average scattering coefficient error of less than 3dB. Additional cases of Southwest China, Hainan Island and the Great Khingan Mountain demonstrate the feasibility of the proposed simulation scheme for various application scenarios.

High-resolution synthetic aperture radar (SAR) as an imaging device becomes more and more like a “camera” at the microwave frequency band. How different objects or object surfaces may visually appear in SAR images becomes an interesting research topic. Inspired by the bidirectional reflectance distribution function (BRDF) models employed in computer graphics (CGs), this article proposes the coherent spatially varying bidirectional scattering distribution function (CSVBSDF) for characterizing the electromagnetic scattering and SAR imaging behavior of surfaces. The CSVBSDF establishes a mapping function from observation parameters and surface local parameters to multidimensional measurements. In this article, CSVBSDF of the randomly rough surface is derived via adapting the integral equation method (IEM) to finite-size pixel cells under the plane wave and tapered wave incidence, respectively. It is then validated against the numerical beam simulation method (BSM) in the SAR image domain. A ground-based rail SAR and a 3-D laser scanner are used to measure the SAR image and the corresponding 3-D geometry of a real ground surface. Surface-local parameters, such as the local slope and roughness, are estimated from the measured 3-D geometry and then fed into the CSVBSDF model to produce a synthetic SAR image. Comparison against the real SAR image preliminarily demonstrates the efficacy of the proposed CSVBSDF model.

Forward modeling of wave scattering and radar imaging mechanisms is the key to information extraction from synthetic aperture radar (SAR) images. Like inverse graphics in the optical domain, an inherently-integrated forward-inverse approach would be promising for SAR advanced information retrieval and target reconstruction. This paper presents such an attempt at inverse graphics for SAR imagery. A differentiable SAR renderer (DSR) is developed, which reformulates the mapping and projection algorithm of the SAR imaging mechanism in the differentiable form of probability maps. First-order gradients of the proposed DSR are then analytically derived, which can be back-propagated from rendered image/silhouette to the target geometry and scattering attributes. A 3D inverse target reconstruction algorithm from SAR images is devised. Several simulation and reconstruction experiments are conducted, including targets with and without background, using synthesized data or real measured inverse SAR (ISAR) data by ground radar. Results demonstrate the efficacy of the proposed DSR and its inverse approach.

Due to the lack of raw data, difficulty in labeling as well as the sensor parameters limitation, few-shot learning in SAR images has become an important research direction. A deep feature transformation method based on differential vector is proposed in this paper to alleviate the feature drift problem in few-shot learning. Firstly, samples generated by a modified Adversarial Auto Encoder (AAE) are introduced as an auxiliary dataset for the few-shot training dataset. Secondly, differential vector is further proposed to alleviate the cross-class-bias between the generated and real data in deep feature space. It is defined as the mean difference of the low-dimensional feature vectors of the real and generated samples, which is considered to be able to describe the transformation direction from generated to real data. Experiments conducted on MST AR dataset demonstrate the feasibility, effectiveness and superiority of proposed method.

In the 2020 Gaofen Challenge on Automated High- Resolution Earth Observation Image Interpretation [1], there is a topic on airplane detection and recognition in optical images. The task is to detect and classify 10 types of civil airplanes, and the difficulty lies in the classification of similar airplanes. We propose a novel airplane detection and recognition method incorporating component detection, which is based on data preprocessing, basic detection network and object head detection. With the experiments between our approach and other detection networks in local training dataset, online validation and test dataset, our method effectively improved the performance of detection and recognition of airplanes in optical images.

This article proposes a MiniSAR imaging algorithm applied to small maneuvering platforms such as multi-rotor UAVs. Due to its load and size limitations, it can only carry low-precision IMU and GPS. These devices are not enough to accurately obtain the motion error of the platform, and thus cannot guarantee the successful imaging of each flight. In this article, an imaging algorithm segment aperture imaging (SAI) algorithm based on time-domain segmentation is designed according to the motion characteristics of a small multi-rotor platform, and the deviation is compensated according to the echo estimation, and then segment stitching is performed to obtain better imaging results. The proposed algorithm is experimentally demonstrated with the Multirotors FUSAR-Ku MiniSAR system.

Land cover classification with SAR images mainly focuses on the utilization of fully polarimetric SAR (PolSAR) images. This paper explores the potential of semantic segmentation of high-resolution single polarimetric (single-pol) SAR and PolSAR images, in particular tailored for the Gaofen-3 (GF-3) sensor. First of all, a unified SAR data preprocessing method is utilized to deal with the L2 format SAR data. Then, an encoder-decoder network based on transfer learning is designed to implement semantic segmentation of GF-3 SAR images. Experiments on single-pol SAR and PolSAR images demonstrate the feasibility of semantic segmentation with high-resolution GF-3 images.

Among the different deep learning-based methods proposed for SAR image despeckling, the main issue seems to construct reliable training data sets. In the statistical-based solution MONet, which assumes square root Gamma distributed speckle in the simulation, the authors showed that despeckling results on actual SAR images are stringently related to the considered training dataset and its statistical distributions. This paper develops realistic simulated data sets for feeding the MONet architecture, including backscattering mechanisms arising in different existing SAR scenarios. We consider a generalized Gaussian coherent scatterer model for SAR correlated clutter simulation for this aim. The use of such simulation has a twofold effect within the considered framework: from one side, it allows generating several noisy patches, used as input data; on the other, it allows including different speckle distributions for different actual SAR scenarios. Results on SAR images show the effectiveness of such simulation.

The interferometric vertical wavenumber (or baseline) has a direct impact on PolinSAR forest height mapping, which must be selected appropriately to acquire optimum inversion performance. In this paper, the key parameters influencing the height estimation precision are considered to simplify the system performance analysis. A PolinSAR performance optimization problem is then established according to the geometrical interpretation of the line coherence model. Finally, the contour map of optimum vertical wavenumber varying with forest height and wave extinction is intuitively provided, from which the system designers can easily determine the optimum baseline for PolinSAR forest height mapping.

3-D reconstruction is a hot topic in remote sensing as well as computer vision. The particularity and complexity of the microwave scattering mechanism bring great challenges to the 3D reconstruction of SAR images, and the applicability of existing methods need to be improved. This study proposes an efficient and explainable point clouds framework for three-dimensional reconstruction of SAR images based on three-dimensional parametric electromagnetic part models. This 3-D SAR reconstruction framework consists of two parts: a feature extraction generative adversarial network and a 3-D reconstruction generative network. The feature extraction generative adversarial network has 5 convolutional layers to extract the features of single SAR image and save them in the form of graph, then input this graph to the 3-D reconstruction generative network and we can get the main shape of the target from a SAR image. This framework effectively reduces the numbers of observation for 3-D reconstruction and make the 3-D reconstruction from single SAR image possible.

The forward representation provides important theoretical basis for SAR image interpretation tasks. The Gaussian coherent scatterer (GGCS) model has been proposed to represent diverse SAR images from both physical and statistical perspectives. It models the SAR image as a coherent summation of a number of Gaussian distributed scattering fields for each resolution cell. The parameters of the Gaussian distribution are assumed to be the same for each scatterer and the scatterer number in a resolution cell is modeled as a random variable. However, the fluctuation of the scatterer number leads to the limitations and inaccuracy of the GGCS model. To solve this problem, this paper proposes a new coherent generative scheme based on an improved GGCS. It consists of a fixed number of random distributed scatterers each of which can have different distribution parameters. Both statistical and correlation characteristics of the scheme are derived. And experiments on simulated data validate the proposed scheme.

Aircraft detection in synthetic aperture radar (SAR) images is a challenging task because of the discreteness, variability, and interference of aircraft scattering features. This article proposes a new hybrid approach of scattering information enhancement (SIE) and an attention pyramid network (APN). It first extracts strong scattering points (SSPs) of aircraft via an adapted Harris-Laplace detector. These SSPs are then clustered into candidate scattering regions by density-based spatial clustering of applications with noise (DBSCAN) and are then modeled with a Gaussian mixture model (GMM). Target scattering clusters are discriminated from background clutter by measuring the Kullback-Leibler divergence (KLD) to the known target templates. These target scattering clusters are enhanced in the preprocessing stage. All the SIE-preprocessed images are then fed into the APN for training and testing. It is composed of the multiscale feature pyramid network (FPN) and the modified convolutional block attention module (CBAM) to cope with the discreteness and variability of aircraft. In addition, focal loss (FL) is adopted to deal with the issue of unbalanced sample distribution and the interference from hard samples. Experiments conducted on the Gaofen-3 and TerraSAR-X data sets demonstrate the effectiveness of the proposed method with an average precision (AP) of 83.25%.

This article presents a robust hand-gesture recognition method via multidimensional feature representation and learning specifically designed for commercial frequency-modulated continuous wave (FMCW) multi-input multi-output (MIMO) millimeter-wave radar. First, the optimal configuration of the radar system parameters for the hand-gesture recognition scenario is investigated and a standard procedure to determine the system configuration is given. Then a moving scattering center model is proposed to represent the 3-D point cloud in the range-Doppler (RD)-angular multidimensional feature space. A scattering point detection and tracking algorithm is presented based on a set of motion constraints in terms of position, velocity, and acceleration. It is derived from the space-time continuity of a nonrigid target. Finally, a lightweight multichannel convolutional neural network (CNN) is designed to learn and classify multidimensional gesture features including radial RD and tangential azimuth-elevation. Extensive experiments are carried out with the developed system and a large data set is obtained to train and test the classifier. The results show that the proposed gesture recognition method can effectively distinguish gestures that are easily confused in the RD domain and achieve robust performances under various conditions.

To date, several different methods for polarimetric synthetic aperture radar (SAR) calibration with distributed targets have been proposed in the literature. The basic assumptions on the distributed target that are used by these methods are almost identical. Their difference is about the assumptions on noise. In this article, the research shows that the subtle difference between the assumptions on noise leads to two different calibration models (CMs), which is the primary cause of the differences between various methods. According to the used CM, the methods in the literature can be categorized into two groups. Because this article focuses on the CMs, thus the optimal estimator in each group is used for comparison so as to exclude the impacts of different parameter estimation algorithms. The results suggest that neither of the optimal estimators is always superior to the other. In practice, we cannot determine which estimator is better, so we recommend using the mean value of the two optimal estimators (i.e.,$ { \widehat {\boldsymbol \varphi }}^{\star } $) for calibration because it was proved to be (at least) better than the worse one. In the research, the signal-to-noise ratio (SNR) was proved to be a proper indicator for assessing whether${ \widehat {\boldsymbol \varphi }}^{\star } $is reliable. Hence, an estimator for the mean SNR is proposed. In this article, some simulation experiments are used to verify some critical conclusions that we have drawn. The practical use of${ \widehat {\boldsymbol \varphi }}^{\star } $and an assessment of its reliability with the estimated mean SNR are illustrated with DLR E-SAR data from the 2006 AgriSAR campaign.

With the development of coding metasurface antennas, here a generalized SAR modality named variable-resolution (VR) SAR is proposed, which involves diverse radiation pattern and inhomogeneous pulse repetition frequency (PRF) to achieve continuously variable azimuth resolution. Optimization of PRF has been conducted by making resolution of some unimportant details decline with no azimuth ambiguities. There are two potential advantages of VR SAR mode: 1) simultaneously offers the higher resolution for the concerned parts of whole scenes and larger image sizes than would otherwise be possible in stripmap and spotlight SAR modes and 2) relaxes the pressure of data storages and brings a high imaging efficiency. The mathematical model of VR SAR system is built up and feasibilities are demonstrated through a series of simulations and an equivalent experiment based on RADARSAT-1 raw data. This VR SAR opens a new venue for earth observation, space-borne remote sensing and related SAR images processing, heading for agile frequencies, beam pattern and beyond.

A zero-shot learning (ZSL) method of automatic target recognition (ATR) in synthetic aperture radar (SAR) image is proposed to address the scenario, where no SAR sample of a particular target is available for training. To learn features of the unseen target, physics-based electromagnetic (EM) simulated images of the target under different azimuth angles are used as the training data instead. The challenge lies in the fact that the simulated image has a distinct but nonessential texture that the real images do not have and, thus, can easily result in an overfitted discriminator network. To overcome this problem, all images are first preprocessed with a nonessential factor suppression step and then fed into a pretrained convolutional neural network for feature extraction. Finally, the feature vector is fed into a trainable fully-connected network for classification. The low-dimensional embedding of feature vectors suggests that the nonessential factor suppression can align the simulated samples with true samples effectively. We propose the max-tolerability principle and averaged margin index for ZSL, which is a useful indicator for selecting optimal classifier. We validated our method on ten-type target recognition task on MSTAR data sets and achieved 91.93% accuracy on nine known targets and 79.08% accuracy on zero-shot target.

A 3-D security imaging experiment at 24 GHz is demonstrated with a 1-D sparse multi-input multi-output (MIMO) array. The MIMO array is an 8 transmission/16 reception arc array to achieve real-aperture imaging along the vertical dimension. It is time-switching multiplexed with a low-cost frequency modulated continuous wave transceiver working at 22-26 GHz. A calibration procedure is proposed to calibrate the channel imbalance across the MIMO array. The experiment is conducted on humans moving on a cart, where we take advantage of the linear motion of humans to form inverse synthetic aperture along the horizontal dimension. To track the motion of humans, a 3-D depth camera is used as an auxiliary sensor to capture the rough position of the target to aid synthetic aperture radar imaging. The back-projection imaging algorithm is implemented on a graphics processing unit for quasi-real-time operations. Finally, experiments are conducted with a real human with concealed objects and a preliminary automatic object detection algorithm based on convolutional neural networks is developed and evaluated on real data.

This article proposes a generalized modeling and simulation approach for correlated synthetic aperture radar (SAR) texture based on the Gaussian coherent scatterer model. It is rooted in the physics-based coherent scatterer assumption where each observation in an SAR image is a coherent sum of multiple underlying Gaussian scatterers. The proposal generalizes existing single-point statistical models by allowing the number of scatterers to be a correlated random field. It can also generate the desired spatial correlation texture by stipulating the structure in both the Gaussian scattered field and the number of scatterers. This generalized model is derived theoretically and then validated by both simulations and experiments with SAR data from actual sensors.

In recent years, the satellite videos have been captured by moving satellite platforms. In contrast to consumers, movies, and common surveillance videos, satellite videos can record the snapshots of city-scale scenes. In a broad fieldof-view of satellite videos, each moving target would be very tiny and usually composed of several pixels in frames. Even worse, the noise signals also exist in the video frames, and the background of the video frames subpixel-level and uneven moving thanks to the motion of satellites. We argue that it is a novel type of computer vision task since previous technologies are unable to detect such tiny moving vehicles efficiently. This paper proposes a novel framework that can identify small moving vehicles in satellite videos. In particular, we offer a novel detecting algorithm based on the local noise modeling. We differentiate the potential vehicle targets from noise patterns by an exponential probability distribution. Subsequently, a multi-morphologicalcue based discrimination strategy is designed to distinguish correct vehicle targets from the existing noises further. Another significant contribution is to introduce a series of evaluation protocols to measure the performance of tiny moving vehicle detection systematically. We annotate satellite videos manually to test our algorithms under different evaluation criterions. The proposed algorithm is also compared with the state-of-the-art baselines, which demonstrates the advantages of our framework over the benchmarks. Besides, the dataset would be downloaded from http://first.authour.github.com.

This paper presents a novel unsupervised polarimetric synthetic aperture radar (PolSAR) image classification method, which incorporates polarimetric image factorization and deep convolutional networks into a principled framework. To implement this idea, we design a convolutional neural network (CNN) with a newly defined loss function which measures the probability distribution distance between the initial distribution maps and CNN predictions. In the proposed method, we firstly execute polarimetric image factorization to generate a dictionary of meaningful atom scatters and their corresponding distribution maps, where the strongest scatters are selected as training samples for CNN. Next, we train the CNN by iteratively optimizing the defined energy function, producing the final distribution maps and classification result. The proposed approach is applied on a real UAVSAR image. Experimental results justify that our approach can effectively classify the PolSAR image in an unsupervised way and produce favorable classification results.

Aiming at improving the classification performance with greatly reduced annotation cost, this paper presents an active deep learning approach for minimally-supervised PolSAR image classification, which integrates active learning and fine-tuning convolutional neural network (CNN) into a principled framework. Starting from a CNN trained using a very limited number of labeled pixels, we iteratively and actively select the most informative candidates for annotation, and incrementally fine-tune the CNN by incorporating the newly annotated pixels. Moreover, to boost the performance and robustness of the proposed method, we employ Markov random field to enforce label smoothness, and data augmentation technique to enlarge the training set. Extensive experiments demonstrated that our approach achieved state-of-the-art classification results with significantly reduced annotation cost.

Target detection is an important application in remote sensing. In this paper, an end-to-end aircraft detection algorithm is proposed for large scene spaceborne synthetic aperture radar (SAR) imagery. Due to the diversity and variability of scattering mechanism, representative features including edge information and depth characteristics are utilized in the algorithm. Firstly, the accurate airport mask is extracted via Otsu algorithm and adaptive identification operator (AIO) with airport morphological features. Secondly, edge-detection based on Canny operator and k-means clustering algorithm are adopted to generate candidate areas. Finally, aircraft targets are discriminated from candidate areas via ResNet-based convolutional neural network (CNN). Experiments are conducted on collected spaceborne SAR imagery, and the results indicate that the proposed algorithm can extract airport area precisely and detect aircraft accurately with low false alarm.

A framework of end-to-end ship detection and recognition for high-resolution Gaofen-3 (GF3) SAR images is proposed. The framework includes three consecutive stages, namely sea-land segmentation, ship detection and discrimination. First, Otsu-based segmentation is used to exclude land areas. Then, adaptive multi-scale constant false alarm rate (CFAR) algorithm is employed to detect candidate ship target pixels. Subsequently, a convolutional neural network (CNN) is designed to filter out false alarms. The CNN is trained by a in-house built GF3 ship dataset, GF3-FUSAR Ships, which is a matchup dataset of SAR and Automatic identification system (AIS).

This paper focuses on the generalization ability of model for SAR automatic target recognition (ATR). An object-based similarity evaluation method for MSTAR datasets is proposed at first to show the relationship between classification accuracy and orientation difference between training and test images. It reveals poor orientation generalization ability of traditional methods for orientation interval larger than 10deg. In order to improve the orientation generalization ability, a novel adversarial autoencoder neural networks (AAN) is proposed in this paper. It learns a code-image-code cyclic network by adversarial training for the purpose of generating new samples at different azimuth angles. The learned orientation predictor and classifier is applied to test samples. Proposed network achieved over 86% classification accuracy on 7-type MSTAR datasets when minimum orientation interval is limited to 25deg, and is about 4% higher than baseline model A-ConvNets under the same condition.

TwinSAR-L (Terrain Wide-swath Interferometric L-band SAR mission) is an innovative space-borne bistatic SAR mission for global dynamics, which will be launched in 2020. This paper investigates ionospheric effects on phase and Faraday rotation of interferometry for TwinSAR-L systems. This ionospheric offset arises from different incident angles along each path in inhomogeneous ionosphere. Plus, the inhomogeneity of ionospheric TEC will cause different range delay and defocusing due to dispersion and azimuth shift. The analysis in this paper will be important for TwinSAR-L mission and potential Tandem-L mission in the future.

Terrain surface classification is probably the most common application of polarimetric SAR (PolSAR) data. Methods for PolSAR terrain classification can be divided into either supervised or unsupervised. In this paper, PolSAR image classification algorithms are reviewed from traditional polarimetric methods such as alpha-H-, Freeman-, Yamaguchi-decomposition, to deep learning, and then a general deep learning algorithm is proposed to PolSAR data classification. The suitability and potential of deep convolutional neural network in supervised terrain classification of PolSAR images has been investigated. The results show that deep learning based method can be used for PolSAR terrain classification.

In this paper, a semi-physical system of SAR jamming based on electromagnetic scattering simulation was designed and constructed by NI and xPC platform. The basic idea is transmitting simulated target scattering echoes to realize jamming for real SAR system. Scattering signal is simulated by our laboratory developed software, and a hardware system was designed to receive SAR signal and transmit simulated echoes. A software radio ADALM-Pluto is used as radar, and jammer adopts NI PXle-1082 and NI PXle-5840. The experimental results show that the designed semiphysical system can reach the goal of jamming by transmitting simulated images. The proposed system can be used to verify SAR jamming and anti-jamming algorithm for researchers.

SAR image due to its nature of coherent imaging manifests both deterministic semantic information and speckle-like statistical textures. It is necessary to have a general representation scheme of the semantic-statistical two-layer hierarchy of SAR image so that semantic and textural information can be separated. Inspired by the correlated clutter simulation method proposed by Bustos et al [1]–[2], this paper studies a semantic-statistical convolution scheme to generate a SAR image from a semantic map. For each terrain type, we estimate the intensity distribution and correlated texture model and then generate textures with correlated clutter. The method is tested on actual SAR images of E-SAR data including urban and forest areas and Flevoland AirSAR data with 15 terrains.

Interferometric synthetic aperture radar (InSAR) uses phase difference of radar echoes, either from multiple passes along the same trajectory or from multiple displaced phase centers on a single pass, to generate interferogram. Scattering correlation in the angular dimension is a critical factor determining the quality of InSAR interferogram. It can be modeled with the angular correlation function (ACF). In this letter, the ACF of a 1-D rough surface under incidence of a tapered wave, namely, a pencil beam, is studied numerically for correlation analysis of InSAR. An analytic ACF is first derived based on the first-order small perturbation method. It is then validated statistically by the method of moment of electromagnetic scattering. Analysis of the ACF simulations indicate that the ACF of backscattering from a randomly rough surface exhibits a shape of sinc function, which depends on tapering parameter g, interferometric incidence angles θ1and θ2. Several numerical simulations of different rough surface spectrums demonstrate that the analytical ACF fits well with numerical results as long as g is the larger several correlation lengths l.

Recently, deep neural networks have received intense interests in polarimetric synthetic aperture radar (PolSAR) image classification. However, its success is subject to the availability of large amounts of annotated data which require great efforts of experienced human annotators. Aiming at improving the classification performance with greatly reduced annotation cost, this paper presents an active deep learning approach for minimally supervised PolSAR image classification, which integrates active learning and fine-tuned convolutional neural network (CNN) into a principled framework. Starting from a CNN trained using a very limited number of labeled pixels, we iteratively and actively select the most informative candidates for annotation, and incrementally fine-tune the CNN by incorporating the newly annotated pixels. Moreover, to boost the performance and robustness of the proposed method, we employ Markov random field (MRF) to enforce class label smoothness, and data augmentation technique to enlarge the training set. We conducted extensive experiments on four real benchmark PolSAR images, and experiments demonstrated that our approach achieved state-of-the-art classification results with significantly reduced annotation cost.

Polarimetric synthetic aperture radar (SAR) data are inevitably contaminated by polarization crosstalk and channel imbalance, which propagate to the error of final remote sensing product. To ensure the successful estimation of forest heights from forthcoming polarimetric SAR interferometry (PolinSAR) campaigns, a critical study on the polarimetric system requirements of PolinSAR for forest height mapping must be carried out. This paper establishes an analytical model for forest height estimation error including dependences on polarimetric system parameters including crosstalk, channel imbalance, and system noise. Simulation analyses are conducted on the real airborne SAR data acquired by the E-SAR system to validate the proposed theoretical error dependence model. We demonstrate that the established error model can be used not only by the system designers as a guidance for setting the polarimetric system requirements of PolinSAR for forest height mapping, but also by the data analyst to correct for systematic bias in the forest height inversion.

With growing needs of the global land cover information, high resolution datasets have been released, such as 30m-resolution GLC30 and NLCD 2011. However, these datasets cannot update in time when the ground-truth changes. In this paper, an automatic algorithm is proposed to generate land cover based on convolutional neural network (CNN) using optical image. A land cover translation framework is designed by utilizing the techniques of the fully convolutional network, and it can provide pixel-to-pixel translation of the image. The experiment is carried out on high resolution optical image sourcing from Google earth, and the results demonstrate that the proposed method is able to generate robust and reasonable prediction of the land cover. By testing at several different areas, this methods achieve the average accuracy at 70.2% for 8 types land cover comparing with NLCD 2011 datasets.

We propose a deep neural networks based method to reconstruct full polarimetric (full-pol) information from single polarimetric (single-pol) SAR data. It consists of two parts: feature extractor which is used to obtain multi-scale multi-layer features of targets in single-pol gray image, and feature translator that converts the geometric features to defined polarimetric feature space. The proposed method is demonstrated on L-band UAVSAR of NASA/JPL images over San Diego, CA, and New Orleans LA, USA. Both qualitative and quantitative results show the reconstructed full-pol images agree well with true full-pol images, the proposed networks have a good spatial robustness. Model-based target decomposition and unsupervised classification can be used directly on constructed full-pol images.

Polarimetric interferometric SAR (PolinSAR) data are contaminated by cross-talk and channel imbalance. To ensure the successful estimation of forest heights from forthcoming PolinSAR campaigns, a critical study on the polarimetric system requirements of PolinSAR for forest height mapping must be carried out. In this paper, a triple-factor analysis of cross-talk, channel imbalance and noise of PolinSAR system is conducted to understand the polarimetric system requirements for PolinSAR forest height mapping. A model relationship between forest height estimation error and polarimetric system parameters is established through theoretical analysis. Meanwhile, the numerical relationship between the two is obtained by artificially adding different system errors to simulated SAR images. The experiment results well validate the correctness of our established model relationship. The polarimetric system requirements of PolinSAR for forest height mapping can be provided for system designers according to our established relationship.

Artificial intelligence such as deep learning has become the dominant approach in computer vision area. It has great potential in improving the performance of SAR automatic target recognition (ATR) as well. In this paper, we present a framework for intelligence SAR ship recognition and a preliminary implementation as well as a demonstration with the ALOS2 data.

A microwave/mmw imaging method for moving objects of non-rigid body using 2D sparse MIMO (multiple-input multiple-output) array is proposed in this paper. To achieve high-resolution 3D image, two orthogonal linear arrays are introduced to form a 2D sparse MIMO array and wide-band signals are transmitted in this method. Subsequently, a space-time trajectory model is introduced to describe the motion of the non-rigid body target. The joint use of segmental and joint-estimation leads to the effective movement compensation of each component. Finally, all the images of components are merged into an image of the whole target and the 3D high-resolution image of the target is reconstructed.

To monitor a space target, 3-D reconstruction from a multiview sequence of the inverse synthetic aperture radar (ISAR) imaging is developed. Scattering of a complex electric-large target, e.g., the ENVISAT satellite model, is numerically calculated, and multiview 2-D ISAR imaging can be simulated. Under the sparse sampling ISAR imaging via compressed sensing, the Kanade-Lucas–Tomasi feature tracker is applied to extraction of target feature points. Then, using the orthographic factorization method, 3-D reconstruction of those feature points is produced. A simple hexagonal frustum is first tested for the feasibility analysis. Two sequences of multiview ISAR imaging, one is the ENVISAT model and another real measurements of a space shuttle, are then presented for 3-D reconstruction. Furthermore, a complex multistructure model of the International Space Station is also studied from multiview ISAR imaging under different sparse sampling rates. All results demonstrate good feasibility of the 3-D reconstruction for those target components, e.g., solar panel and antenna.

Multi-station inverse synthetic aperture radar (MS-ISAR) imaging mode for on-orbit space target is presented. The MS-ISAR network produces bistatic ISAR images from different radar systems in different locations, and may retrieve more information of on-orbit space target, especially, for applications of target detection and recognition, as well as three dimensional (3-D) reconstruction. Numerical scattering/MS-ISAR imaging of a complex electrically-large target, which is different from simple point-scatterers, are simulated to demonstrate the facilities of multi-station networks, and an example of the Envisat on real orbits is presented.

Compared to quad polarimetry (quad-pol), compact polarimetry (compact-pol) can reduce the system complexity and data volume but sacrifice the retrievable information content. Reconstruction of pseudo quad-pol data from compact-pol has been proposed mostly based on iterative algorithms which make use of the empirically parameterized model with the assumption of reflection symmetry. In this paper, a systemic inverse problem model for quad-pol reconstruction is formulated by developing the linear relationships between the three compact-pol and quad-pol covariance matrices. We then developed a novel Wishart-Bayesian regularized method to solve the inverse problem. The method is verified with Flevoland AirSAR data.

Zero-shot learning (ZSL) is of critical importance for practical synthetic aperture radar (SAR) automatic target recognition (ATR) as training samples are not always available for all targets and all observation configurations. We propose a novel generative-based deep neural network framework for ZSL of SAR ATR. The key component of the framework is a generative deconvolutional neural network referred to as generator. It learns a faithful hierarchical representation of known targets while automatically constructing a continuous SAR target feature space spanned by orientation-invariant features and orientation angle. It is then used as a reference to design and initialize an interpreter convolutional neural network, which is inversely symmetric to the generator network. The interpreter network is then trained to map any input SAR image, including those of unseen targets, into the target feature space. In a preliminary experiment with the Moving and Stationary Target Acquisition and Recognition data set, seven targets are used in the training of generator and interpreter networks. Then, the eighth target is used to test the interpreter, where it is correctly mapped to the reasonable spot spanned by the previous seven targets and its orientation can also be estimated.

This letter closes a special stream consisting of selected papers from the fifth Asia-Pacific Conference on Synthetic Aperture Radar in 2015 (APSAR 2015). The latest research results and outcomes from APSAR 2015, particularly on the synthetic aperture radar (SAR) systems/subsystems design, data processing techniques, and various SAR applications in remote sensing, are summarized and presented. All these results represent the recent advances in SAR remote sensing. Hopefully, this letter can provide some references for SAR researchers/engineers and stimulate the future development of SAR technology for remote sensing.

Among high-squint high-resolution (HSHR) synthetic aperture radar imaging algorithms, nonlinear chirp scaling algorithm (NLCSA) and its extensions, such as extended NLCSA (ENLCSA), have a common drawback in that they all neglect the spatial variations of linear range migration (LRM) and Doppler centroid, and thus, only targets in a specific central slant range plane can be strictly focused. In this letter, we show that by using a new NLCSA, targets can be focused in the ground plane under HSHR conditions. Based on a more accurate 2-D spectrum, the new NLCSA outperforms the ENLCSA by introducing a new range–Doppler domain interpolation to correct residual range migration and a new perturbation function to remove the dependence of Doppler phase on azimuth. The coefficients of the new perturbation function are numerically calculated and then smoothed by polynomial fitting. Though the outputs of the numerical calculation are somewhat unstable at the current stage, it has been demonstrated to perform better than the algorithms neglecting the spatial variations of LRM and Doppler centroid, such as the ENLCSA, by point target simulations.

Layover scatterers of tall building structures can be separated by synthetic aperture radar tomography (SAR-tomo). An iterative reweighted L1 minimization (IRL1) has been applied to enhance the sparsity in a tomographic inversion, where the basis pursuit (BP) technique was adopted to search for the solution. However, the IRL1 with BP is highly time-consuming, which may prevent its real application to large-scale data sets. In this letter, we propose the iterative reweighted alternating direction method of multipliers (IR-ADMM) for fast SAR-tomo imaging. We demonstrate and validate the enhanced sparsity and fast convergence of our IR-ADMM algorithm with experiments using both simulated data and TerraSAR-X Stripmap images of tall urban buildings. The experimental results show that compared with conventional IR-BP, the IR-ADMM greatly reduces the computation time without substantial performance degradation.

Compact polarimetry (compact-pol), as an effective polarization system, can reduce the system complexity and data volume in comparison with quad polarimetry (quad-pol). Reconstruction of quad-pol data from compact-pol has been discussed mostly based on iterative algorithms which make use of the empirically parameterized model with the assumption of reflection symmetry of the scatterer. In this letter, a linear relationship between the compact-pol and quad-pol is first derived, and then the Wishart-Bayesian regularized inverse algorithm is developed to reconstruct pseudo quad-pol data from compact-pol. Such problem is solved using the efficient alternating direction method of multipliers to recover the pseudo quad-pol covariance matrix. The reconstruction performance is evaluated by coherence index, in comparison with existing methods.

This paper reformulates the problem of polarimetric incoherent target decomposition as a general image factorization which aims to simultaneously estimate a dictionary of meaningful atom scatterers and their corresponding spatial distribution maps. Both model-based and eigenanalysis-based decompositions can be seen as special cases of image factorization under specific constraints. The inverse problem of image factorization can be converted to an equivalent nonnegative matrix factorization (NMF) problem via redundant coding. It enables a wide range of NMF algorithms with various regularizations to be directly applicable to polarimetric image analysis. The advantage of the proposed image factorization is demonstrated on both synthesized and real data. It also shows that extended applications such as speckle reduction and classification can benefit from the proposed image factorization.

In high-resolution synthetic aperture radar (SAR) images, the spatial pattern of a homogenous region conveys rich information regarding the specific scatterers being imaged. Analytical modelings of both scattering and imaging processes are critical to better interpret a specific scatterer's appearances in SAR imagery and thereafter quantitatively retrieve its physical parameters. Small-scale rough surface scattering represents one of the mostly common scattering mechanisms and yet its SAR image characteristics have not been well studied. This becomes more critical in millimeter-wave/terahertz regime as smooth surfaces would become slightly rough under millimeter/submillimeter wavelengths. In this paper, we recast the small perturbation method (SPM) approximation of rough surface scattering under the deterministic finite-length surface condition. By ignoring evanescent waves, a simplified SPM solution for rough facet is derived as well as the analytical form of its SAR image under the conventional setup. Then, we reformulate the scattering imaging process of rough facet as a signal processing chain that further reveals the underlying mechanism of rough surface as imaged by radar. The proposed method is extensively validated against the method of moments in terms of both scattering coefficients and imaging patterns. It is found that under the conventional SPM validity condition, the error of scattering coefficients is less than 1.5 dB, while the correlation between SPM-calculated and MoM-calcualted SAR images is larger than 0.9. The proposed model of rough facet SAR image could serve as the theoretical basis for parameter inversion and surface reconstruction. Preliminary cases of the inversion of rough surface spectrum using both simulated and real SAR image patches are presented.

Following the great success of deep convolutional neural networks (CNNs) in computer vision, this paper proposes a complex-valued CNN (CV-CNN) specifically for synthetic aperture radar (SAR) image interpretation. It utilizes both amplitude and phase information of complex SAR imagery. All elements of CNN including input-output layer, convolution layer, activation function, and pooling layer are extended to the complex domain. Moreover, a complex backpropagation algorithm based on stochastic gradient descent is derived for CV-CNN training. The proposed CV-CNN is then tested on the typical polarimetric SAR image classification task which classifies each pixel into known terrain types via supervised training. Experiments with the benchmark data sets of Flevoland and Oberpfaffenhofen show that the classification error can be further reduced if employing CV-CNN instead of conventional real-valued CNN with the same degrees of freedom. The performance of CV-CNN is comparable to that of existing state-of-the-art methods in terms of overall classification accuracy.

This paper presents a three-dimensional (3-D) target reconstruction characterized by its geometric components, i.e. canonical scattering objects. The target is decomposed into some primitive geometries, which are extracted to represent and characterize scattering features. The object parameters are estimated in the frequency domain based on the least square method, and all estimated objects uses the model-based reasoning rules to reconstruct 3-D target. The simulation of a simplified tank model is used to validate the feature extraction and target reconstruction.

Inspired by nonlocal filtering, this paper proposes a saliency detector based on pattern recurrence. In visual attention and saliency detection framework, people are extracting patches that are not redundant, which would more likely to attracting attention. Hence, target detection or saliency detection in SAR image could also be done using the dissimilarity as an indicator of saliency, meaning that the interesting target or saliency target is different from background around it. Similarity of two pixels can be defined together with their local neighbors, then calculate the cross-correlation of two normalized patches. To analyze results better, a normalized version of cross-correlation is used. Experimental results on SAR image are shown to test the effectiveness of the proposed method. The experimental results compared with CFAR on SAR images also prove the effectiveness in saliency detection on SAR images.

Using numerical scattering simulation of the BART (bidirectional analytic ray tracing method), an angular sequence of high-resolution 2-D ISAR (inverse synthetic aperture radar) images of a space target is produced. The Kanade-Lucas-Tomasi (KLT) feature tracker is then adopted for extracting feature points and matching all angularly consecutive ISAR images. 3-D positions of those featured points can be retrieved by the orthographic factorization method (OFM). As a test, a simple hexagonal frustum is used for validation and analysis. Then, two ISAR imaging sequences, one is the simulation of the Envisat satellite model and another real measurements of the space shuttle, are presented to demonstrate 3-D shape information. It shows good feasibility for 3-D status evaluation and pointing control of solar array of a satellites in orbit.

Multi-signal compressed sensing with total variation (MTV-CS) is developed for tomographic inversion of building structure. Incorporating with prior information of the building object, some particularly aligned pixels are combined via the minimization of the object function, as indicated by total variation regularization. A numerical simulation of scattering and SAR imaging of the buildings and the TerraSAR-X imaging data are applied for MTV-CS inversion.

To explore anisotropic information embedded among sub-aperture SAR images, we propose polarimetric sub-aperture singular value decomposition, where polarimetric and anisotropic features are simultaneously decomposed. The decomposed singular values and left singular vectors are equivalent to eigen-analysis-based polarimetric target decomposition, while the right singular vectors give the corresponding anisotropic feature vectors. A physics-based parameterization is proposed for anisotropic pattern, where two new parameters, namely, compactness and directivity, are proposed. Both simulation results and real SAR image analyses demonstrate that these new anisotropic parameters can identify specific types of scatterers.

Rapid advances in synthetic aperture radar (SAR) technologies have brought challenges in image interpretation toward the development of new Earth observation applications. Both novel scattering-imaging models and intelligent inversion techniques are required for advanced information retrieval and interpretation of high-resolution multidimension and multimode SAR data. As an example, scene reconstruction attempts to transfer the SAR image to human-understandable representation of man-made targets and natural environment. In this letter, a framework for scene reconstruction is outlined. It includes three key elements: a dictionary of parametric scatterer model, a method for scatterer recognition and parameter estimation, and a method for target reconstruction. A preliminary case is presented, where a simulated 3-D SAR image of a simple target is successfully reconstructed to a solid geometry. It uses a novel surface extension method to connect isolated scatterers to form a complete target geometry.

Deep convolutional neural networks have achieved great success in computer vision and many other areas. They automatically extract translational-invariant spatial features and integrate with neural network-based classifier. This letter investigates the suitability and potential of deep convolutional neural network in supervised classification of polarimetric synthetic aperture radar (POLSAR) images. The multilooked POLSAR data in the format of coherency or covariance matrix is first converted into a normalized 6-D real feature vector. The six-channel real image is then fed into a four-layer convolutional neural network tailored for POLSAR classification. With two cascaded convolutional layers, the designed deep neural network can automatically learn hierarchical polarimetric spatial features from the data. Two experiments are presented using the AIRSAR data of San Francisco, CA, and Flevoland, The Netherlands. Classification result of the San Francisco case shows that slant built-up areas, which are conventionally mixed with vegetated area in polarimetric feature space, can now be successfully distinguished after taking into account spatial features. Quantitative analysis with respect to ground truth information available for the Flevoland test site shows that the proposed method achieves an accuracy of 92.46% in classifying the considered 15 classes. Such results are comparable with the state of the art.

The algorithm of synthetic aperture radar automatic target recognition (SAR-ATR) is generally composed of the extraction of a set of features that transform the raw input into a representation, followed by a trainable classifier. The feature extractor is often hand designed with domain knowledge and can significantly impact the classification accuracy. By automatically learning hierarchies of features from massive training data, deep convolutional networks (ConvNets) recently have obtained state-of-the-art results in many computer vision and speech recognition tasks. However, when ConvNets was directly applied to SAR-ATR, it yielded severe overfitting due to limited training images. To reduce the number of free parameters, we present a new all-convolutional networks (A-ConvNets), which only consists of sparsely connected layers, without fully connected layers being used. Experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) benchmark data set illustrate that A-ConvNets can achieve an average accuracy of 99% on classification of ten-class targets and is significantly superior to the traditional ConvNets on the classification of target configuration and version variants.

An iterative alternating direction method of multipliers (ADMM) is proposed for inverse finite-element–boundary-integral (FE–BI) problem with total variation (TV) regularization. The inverse FE–BI fits to a wide class of penetrable sensing applications, where this study specifically targets the problem of radio tomography of asteroid interior structure using orbiting spacecraft. The TV regularizer enforces sparsity on the gradient of reconstructed permittivity, which agrees well with the “piecewise constant” reality of “rocks embedded in soil” scenario and, meanwhile, addresses the inherent ill-posedness. For large-scale asteroid problems, the distributed ADMM algorithm is adapted to solve the linear TV inversion at each iteration. The 2-D inversion is validated with the Fresnel Institute measurement data. Simulated cases of asteroid internal imaging are also presented. The proposed iterative ADMM can be also applied to similar penetrable imaging applications.

In the booming era of high-resolution synthetic aperture radar (SAR) technology, SAR advanced information retrieval is critical for effective utilization of huge-volume SAR data. One important aspect of high-resolution SAR interpretation is to explore the anisotropic and dispersive information embedded among subaperture and subband SAR images. This paper formulates the polarimetric subaperture analysis as a singular-value decomposition problem, where polarimetric and anisotropic features can be simultaneously decomposed. The decomposed singular values and left singular vectors are equivalent to eigenanalysis-based polarimetric target decomposition, whereas the right singular vectors give the corresponding anisotropic feature vectors. A physics-based parameterization is proposed for anisotropic patterns, where two anisotropic entropy parameters, namely, compactness and directivity, are proposed. Both simulation results and real SAR image analyses demonstrate that these proposed anisotropic entropies can effectively identify specific types of scatterers depending on their geometric scale, curvature, and form of spatial distribution. The proposed anisotropic entropies could be applied to single- and dual-polarization high-resolution SAR data as well.

In this paper, a new All-Convolutional Networks (A-ConvNets) is proposed and applied to Moving and Stationary Target Acquisition and Recognition (MSTAR) data. Conventional deep learning algorithms, especially the deep convolutional networks (ConvNets) have achieved many success state-of-art results. However, directly applying ConvNets to SAR data will yield severe overfitting because of limited data availability. The proposed A-ConvNets can significantly reduce the number of free parameters and the degree of overfitting. Average accuracy of 99.1% on classification of 10-class targets was obtained by applying A-ConvNets to MSTAR datasets.

Man-made building objects mostly with vertical wall structures may present distinct scattering patterns, e.g., wall/roof upfront scattering, wall-ground double scattering, etc., along the range dimension in high-resolution synthetic aperture radar (SAR) images. In this letter, a 1-D detector, referred to as the “range detector,” is presented for building detection, which operates only along the range direction. Experiments show that this range detector can effectively detect and extract the footprint of the illuminated wall of a cuboid building, with which the outline of the building image can be captured by marching the footprint toward radar. This approach is applied to an airborne Pi-SAR image of Sendai, Japan, and more than 80% of the buildings can be identified. The building height and length are also estimated, and the errors are found around 4-5 m based on optical image.

Spaceborne high-resolution synthetic aperture radar (SAR) is a potential powerful tool for rainfall pattern and intensity observations over the sea surface. However, many interesting rain-related phenomena revealed by SAR images are still not fully understood due to poor theoretical modeling of the rain–wind–wave interactions. This paper attempts to develop a physics-based radiative transfer model to capture the scattering behavior of rainfall over a rough sea surface. Raindrops are modeled as Rayleigh scattering nonspherical particles, whereas the rain-induced rough surface is described by the Log-Gaussian ring-wave spectrum. The model is validated against both empirical models and measurements. A case study of collocated Envisat ASAR data and NEXRAD rain data is presented to demonstrate the performance of the newly developed model. Finally, numerical simulation results suggest that rain-related scattering becomes significant as compared with wind-related scattering when the frequency is above C-band, whereas the raindrop volumetric scattering becomes significant above X-band.

A theoretical model of stratified media with non-spherical scatterers and rough interfaces for numerical simulation of polarimetric radar pulse echoes is developed. The time domain Mueller matrix solution derived from vector radiative transfer formulations contains seven scattering mechanisms of the stratified media: surface scattering from the rough top and bottom interfaces, volumetric scattering from random non-spherical scatterers, and their multi-interactions. Temporal characteristics and image features of the polarimetric echo profile as functional dependences on the model parameters, such as the layered structure, thickness and dielectric properties etc., are numerically simulated. Polarimetric pulse echoes may reveal internal structure and other useful information of the layered media, and demonstrate a potential new way to explore the lunar regolith layers and monitor the landslides, etc.

To monitor debris flows and landslides, geologic surveying has been usually implemented to ascertain where these natural hazards might happen. These traditional observations at discrete sites are very restrictive in both temporal and spatial scales, and cannot make accurate and timely decision for early warning of geologic disaster occurrences. In this letter, very high frequency (VHF; ~100 MHz) pulsed radar is proposed as a monitoring tool to probe the layering land media. Due to large penetration depth of VHF radar on the order of tens of meters, radar echoes can detect the change of water content underneath ground surface, which is an essential stimulator to cause the debris flow and landslides. A model of layering land media embedded by random scatterers (stone or water) with randomly rough interfaces is presented, and polarimetric radar range profiles from underground structures under different situations are numerically simulated. Results show that distinct features in radar range profiles can be directly attributed to underground water content change and/or water distribution. The proposed VHF radar seems promising for early warning of geologic hazards. The differences of radar images between the normal day and warning days, e.g., after severe storm, can be used to predict potential occurrence of debris flow or landslides.

Numerical study of radar echoes from the targets in environmental clutters has been of great interest in many applications. In this paper, the bidirectional analytic ray tracing (BART) method for composite scattering from three-dimensional (3D) electrically large complex target above a randomly rough surface is reported. Analytic tracing of polygon ray tubes in bidirectional tracing is developed to precisely calculate the illumination and shadowing of facets, which exempt large patches of the target from any finer meshing. It significantly reduces the complexity relevant to the target electric-size. Numerical examples of angularly composite scattering from a three-dimensional electrically large, e.g., a ship-like target over a randomly rough surface are presented and discussed.

Based on the statistics of the lunar cratered terrain, e.g. population, dimension and shape of craters, the terrain feature of cratered lunar surface is numerically generated. According to inhomogeneous distribution of the lunar surface slope, the triangulated irregular network is employed to make the digital elevation of lunar surface model. The Kirchhoff approximation of rough surface scattering is then applied to simulation of lunar surface scattering. The synthetic aperture radar (SAR) image for comprehensive cratered lunar surfaces is numerically generated. Making use of the digital elevation and Clementine UVVIS data at Apollo 15 landing site as the ground truth, an SAR image at Apollo 15 landing site is simulated.

Employing three-dimensional mapping and projection algorithm (MPA), imaging simulation of bistatic SAR (BISAR) observation over complex scenario is developed. Based on the explicit expression of point target response of stripmap BISAR imaging, raw data is efficiently generated from the scattering map pre-calculated by MPA. Some examples of BISAR image simulation are studied. Polarimetric characteristics of BISAR image are then discussed. A transform of unified bistatic polar bases for BISAR image is proposed. Analysis of simulated images shows that the redefined parameters by the unified bistatic polar bases transform well describe different scattering mechanisms in BISAR imaging. It provides a primary tool for BISAR image interpretation and terrain classification.

Employing the 3-D mapping and projection algorithm (MPA), an imaging simulation of bistatic synthetic aperture radar (BISAR) observation over a complex scenario is developed. Based on the explicit expression of the point target response of stripmap BISAR imaging, raw data are efficiently generated from the scattering map precalculated by MPA. Some examples of BISAR image simulation are studied. The polarimetric characteristics of a BISAR image are then discussed. It is found that some typical polarimetric parameters such as Cloude's alpha, beta and gamma and might become unable to describe the scattering mechanism under bistatic observation. A transform of unified bistatic polar bases for a BISAR image is proposed. The parameters alpha, beta and gamma and are modified to retain the property of orientation independence in the bistatic circumstance. Analysis of simulated images shows that the redefined alpha, beta and gamma and after the unified bistatic polar bases transform well describe different scattering mechanisms in BISAR imaging. It provides a primary tool for BISAR image interpretation and terrain classification.

In this paper, an approach to the automatic reconstruction of 3D simple building objects from multi-aspect metric-resolution SAR images is proposed. The edge detector of constant false alarm rate (CFAR) and a parallel Hough transform technique are first employed to extract the parallelogram-like image of the building walls in SAR image. A set of probability density functions is presented to describe the extracted random wall-images and their multi-aspect coherence. Then the maximum-likelihood estimation of object is derived from its multi-aspect object-images. A hybrid priority criterion is defined to evaluate the reliability of reconstruction result, based on which, an automatic reconstruction algorithm is further devised to match object-images of different aspects and finally reconstruct the building objects. Four-aspect Pi-SAR images over Sendai, Japan are adopted for reconstruction. The results show the fidelity of the whole process chain and the feasibility of 3D objects automatic reconstruction from multi-aspect SAR images.

Reconstruction of 3-D objects from multiaspect high- resolution synthetic aperture radar (SAR) images is of great importance for SAR technology applications. In this paper, simple building objects are modeled as cuboids, and an approach for automatic reconstruction of 3-D building objects from multiaspect SAR images in meter resolution is developed. The edge detector of constant false alarm rate and a Hough transform technique for parallel line segment pairs are first employed to extract the parallelogram-like image of the building walls in SAR images. A set of probability density functions is presented to describe the object images and their multiaspect coherence. The maximum-likelihood estimation of an object is then derived from its multiaspect object images. A hybrid priority criterion is defined to evaluate the reliability of the reconstruction result. An automatic reconstruction algorithm is further developed to match object images of different aspects and, finally, to reconstruct the building objects. Besides, an iterative method is proposed for the coregistration of multiaspect building images. Four-aspect simulated images of a virtual scene and four-aspect Pi-SAR images over the campus of Tohoku University, Japan, are investigated. Reconstruction of building objects from their multiaspect images shows the fidelity of the whole process chain and the feasibility of 3-D objects automatic reconstruction from multiaspect SAR images. At last, a practical application that is based on spaceborne meter-resolution SAR is proposed.

Deorienation theory of polarimetric scattering targets is developed, which transforms the scattering vector of spatially oriented targets into a certain status with minimization of cross polarization. A new set of the parameters u,v,w,Psiis defined to describe and classify different terrain surfaces. Based on the vector radiative transfer (VRT) model of non-spherical particles above a rough surface, numerical simulations illustrate the parameters u,v,w,Psiand the entropy H. These parameters are applied to the unsupervised classification in polarimetric images. The terrain surfaces of polarimetric SIR-C and airborne SAR images are classified and orientation-analyzed.

A novel fast algorithm of polarimetric image simulation for SAR observation over comprehensive terrain scene is developed based on the mapping and projection principles. It incorporates penetrable and impenetrable objects, volumetric and surface scatterers in the imaging space with the extinction, attenuation, shadowing and multiple scattering effects. Scattering of the vegetation canopy is modeled as a layer of random non-spherical particles by using the vector radiative transfer model. Scattering from the ground surface and building objects is calculated by using the IEM rough surface model. As an example, the polarimetric SAR images for a virtual terrain scene, composed by tree canopies, farmland, buildings, rough land surface, hills and rivers, are simulated.

Under a polarized pulse wave incidence, the temporal Mueller matrix solution from vector radiative transfer (VRT) equation for a layer of nonspherical particles above randomly rough surface is constructed. The system response function based on the Mueller matrix solution is developed, which takes into account the scattering intensity matrix of the canopy, attenuation coefficient matrix through the canopy, scattering intensity matrix of underlying ground surface, and echoes time delay. This system response model preserves consistence with the Mueller matrix solution. To evaluate the system response function from the wave profiles of the received pulse echoes, an adaptive nonlinear estimation method (ALM) is proposed. When the pulse echoes are received, it yields the system response function, i.e., four system parameters. These system parameters are used to invert multiparameters of the vegetation canopy and underlying rough surface, which include the canopy depth, scatterers size, orientation, density and dielectric constant, and the surface roughness and dielectric constant. Numerical examples show good performance of our method as a tractable approach for multiparameters inversion. Potential application and some issues of multiparameters inversion are discussed. An envisaged sensor and platform for practical realization is proposed

A novel approach to polarimetric image simulation for synthetic aperture radar (SAR) observation over comprehensive terrain scenes is developed based on mapping and projection principles. It incorporates penetrable and impenetrable objects, volumetric and surface scatterers in the imaging space with extinction, attenuation, shadowing, and multiple-scattering effects. Scattering of the vegetation canopy is modeled as a layer of random nonspherical particles by using the vector radiative transfer model, and scattering from the ground surface and building objects is calculated by using the integral equation method. As an example, polarimetric SAR images at L-band and C-band and the different spatial resolutions for a virtual terrain scene composed of tree canopies, farmland, buildings, rough land surface, hills, and rivers are simulated. The imaging simulation results demonstrate the feasibility of the mapping and projection approach and the potential utilities of SAR imaging simulation

Deorientation theory of polarimetric scattering targets is presented. Using a transformation of the target scattering vector, the target orientation is turned to a certain fixed state and polarimetric scattering of the transformed scattering vector shows the prominence of the generic characteristics of the target. A new set of parameters u, v, w, /spl psi/ is defined based on a deorientation of the target scattering vector. Numerical simulation of polarimetric scattering of nonspherical particles illustrates the meanings of the parameters u, v, w, /spl psi/ and the entropy H. An unsupervised classification scheme of the terrain surfaces is developed, which classifies the terrain surfaces using the set of u., v, H, and analyzes the orientation distribution of each class based on deorientation angle /spl psi/. As examples, a SIR-C polarimetric image over China's Guangdong Hui-Yang district is classified into eight classes and a AirSAR polarimetric image over Canada's Boreal district is orientation-analyzed using our approach of deorientation and four parameters u, v, /spl psi/, and H.