Hyperspectral image (HSI) plays a vital role in various fields such as agriculture and environmental monitoring. However, due to the expensive acquisition cost, the number of hyperspectral images is limited, degenerating the performance of downstream tasks. Although some recent studies have attempted to employ diffusion models to synthesize HSIs, they still struggle with the scarcity of HSIs, affecting the reliability and diversity of the generated images. Some studies propose to incorporate multi-modal data to enhance spatial diversity, but spectral fidelity cannot be ensured. In addition, existing HSI synthesis models are typically uncontrollable or only support single-condition control, limiting their ability to generate accurate and reliable HSIs. To alleviate these issues, we propose HSIGene, a novel HSI generation foundation model which is based on latent diffusion and supports multi-condition control, allowing for more precise and reliable HSI generation. To enhance the spatial diversity of the training data while preserving spectral fidelity, we propose a new data augmentation method based on spatial super-resolution, in which HSIs are upscaled first, and thus abundant training patches could be obtained by cropping the high-resolution HSIs. In addition, to improve the perceptual quality of the augmented data, we introduce a novel two-stage HSI super-resolution framework, which first applies RGB bands super-resolution and then utilizes our proposed Rectangular Guided Attention Network (RGAN) for guided HSI super-resolution. Experiments demonstrate that the proposed model is capable of generating a vast quantity of realistic HSIs for downstream tasks such as denoising and super-resolution.

Class Incremental Learning (CIL) aims to learn new concepts consistently from a data stream without forgetting. Unlike typical CIL methods which need to learn a model from scratch, pre-trained model (PTM) can easily adapt to a new task with fine-tuning. However, existing PTM-based CIL methods fail to achieve a trade-off between performance and computational expenditure, i.e., they either adopt the same parameter space so that leading catastrophic forgetting, or expand a new branch for each task but adding more computational cost. To this end, we propose MetrIc Learning with Expandable Subspace (Miles) to harness the prior information within pre-trained knowledge, thereby orchestrating an efficient expansion of the parameter space through guided optimization. Specifically, it decouples the learnable modules with the pre-trained model, exploiting prior information from intermediate features of the backbone network to enable more flexible parameter expansion. Then, a central loss is adopted to guide the new category to cluster towards the corresponding prototype in the new task subspace while incorporating an auxiliary distance regularization term to maintain metric equilibrium across tasks. Extensive experiments on six benchmark datasets demonstrate that Miles achieves state-of-the-art performance in various CIL settings.

Class incremental learning (CIL) is the key to achieving adaptive vision intelligence, and one of the main streams for CIL is network expansion (NE). However, state-of-the-art (SOTA) methods usually suffer from feature diffusion, growing parameters, feature confusion, and classifier bias. In view of this, a novel dynamic structure dubbed as recurrent NE (RNE) is proposed by establishing connections among task experts. Specifically, the previous task experts transfer features sequentially through a shared module and the new task expert makes adjustments based on received features rather than reextracted ones, thereby focusing more on the key area and avoiding feature diffusion. Furthermore, the RNE is compressed by replacing additional task experts with lightened ones, in order to significantly reduce the number of parameters while keeping the performance almost unaltered. In addition, feature confusion is alleviated by a decoupled classifier and classifier bias is corrected by pseudo-feature generation. Extensive experiments on four widely adopted benchmark datasets, i.e., CIFAR-100, ImageNet-100, Food-101, and ImageNet-1K, have demonstrated that RNE achieves SOTA performance in both ordinary and challenging CIL settings.

Current remote sensing semantic segmentation methods are mostly built on the close-set assumption, meaning that the model can only recognize pre-defined categories that exist in the training set. However, in practical Earth observation, there are countless new categories, and manual annotation is impractical. To address this challenge, we first attempt to introduce training-free1open-vocabulary semantic segmentation (OVSS) into the remote sensing context. However, due to the sensitivity of remote sensing images to low-resolution features, distorted target shapes and ill-fitting boundaries are exhibited in the prediction mask. To tackle these issues, we propose a simple and universal upsampler, i.e. SimFeatUp, to restore lost spatial information of deep features. Specifically, SimFeatUp only needs to learn from a few unlabeled images, and can upsample arbitrary remote sensing image features. Furthermore, based on the observation of the abnormal response

Remote sensing image object detection (RSIOD) aims to identify and locate specific objects within satellite or aerial imagery. However, there is a scarcity of labeled data in current RSIOD datasets, which significantly limits the performance of current detection algorithms. Although existing techniques, e.g., data augmentation and semi-supervised learning, can mitigate this scarcity issue to some extent, they are heavily dependent on high-quality labeled data and perform worse in rare object classes. To address this issue, this paper proposes a layout-controllable diffusion generative model (i.e. AeroGen) tailored for RSIOD. To our knowledge, AeroGen is the first model to simultaneously support horizontal and rotated bounding box condition generation, thus enabling the generation of high-quality synthetic images that meet specific layout and object category requirements. Additionally, we propose an end-to-end data augmentation framework that integrates a diversity-conditioned generator and a filtering mechanism to enhance both the diversity and quality of generated data. Experimental results demonstrate that the synthetic data produced by our method are of high quality and diversity. Furthermore, the synthetic RSIOD data can significantly improve the detection performance of existing RSIOD models, i.e., the mAP metrics on DIOR, DIOR-R, and HRSC datasets are improved by 3.7%, 4.3%, and 2.43%, respectively. The code is available at here.

Recently, road graph extraction has garnered increasing attention due to its crucial role in autonomous driving, navigation, etc. However, accurately and efficiently extracting road graphs remains a persistent challenge, primarily due to the severe scarcity of labeled data. To address this limitation, we collect a global-scale satellite road graph extraction dataset, i.e. Global-Scale dataset. Specifically, the Global-Scale dataset is ∼ 20× larger than the largest existing public road extraction dataset and spans over 13,800 km2globally. Additionally, we develop a novel road graph extraction model, i.e. SAM-Road++, which adopts a node-guided resampling method to alleviate the mismatch issue between training and inference in SAM-Road [17], a pioneering state-of-the-art road graph extraction model. Furthermore, we propose a simple yet effective "extended-line" strategy in SAM-Road++ to mitigate the occlusion issue on the road. Extensive experiments demonstrate the validity of the collected Global-Scale dataset and the proposed SAM-Road++ method, particularly highlighting its superior predictive power in unseen regions. The dataset and code are available at https://github.com/earth-insights/samroadplus.

Because of the constraints of observation conditions, it is difficult to obtain a large amount of measured data for real targets in the inverse synthetic aperture radar (ISAR) system. Existing deep networks usually use the simulated data of random points for training, which will lead to the degradation of the imaging performance of measured data when the distribution of measured data is different from that of simulated data, i.e., poor generalization performance. A high-resolution ISAR imaging method based on Swin Transformer-based deep reinforcement learning (SwinRL) is proposed to address this problem. The 2-D alternating direction method of multipliers (ADMM) is modeled as a sequential decision problem in this method. The internal adjustable parameters are modeled as actions, and the Swin Transformer is used as the backbone network of the policy network and value network. The optimization of the actions, i.e., the internal adjustable parameters of the 2-D ADMM algorithm, is then guided through network training in a reinforcement learning framework. After that, the trained agent can automatically give optimal internal parameters according to different input data, and then well-focused imaging results can be obtained by executing a 2-D ADMM algorithm with optimal parameters. Finally, experimental results based on simulated and measured data show the performance priority of the proposed method compared to existing deep unrolling networks with fixed parameters.

Limited by observational constraints and non-cooperative targets, SAR images often suffer from extreme scarcity of training samples for certain classes, which severely degrades the detection performance and leads to the challenging issue of few-shot SAR object detection. To address this, we propose a Few-Shot SAR object Detection model with Dynamic perception RPN and category knowledge Decoder, dubbed as FSD3. Specifically, a dynamic perception RPN is employed to infuse global semantics of support classes into local features of query feature maps, thereby enhancing the recall rate of novel-class targets under few-shot conditions. Subsequently, a category knowledge decoder is designed, which automatically decodes aggregated features by utilizing query features as prompts and effectively mitigates the feature-label mismatch phenomenon. Finally, a dual-constrained detection head is proposed, which leverages the large margin cosine loss and center calibration loss to explicitly achieve inter-class separation and intra-class compactness, thereby achieving high-precision and robust classification in few-shot scenarios. Experimental results on the public SAR-AIRCraft-1.0 dataset and our custom MSTAR-Fewshot dataset have demonstrated that the proposed model achieves superior few-shot object detection performance across various experimental configurations.

The collaborated observation of multi-static ISAR can provide comprehensive information for air/space target classification. However, under limited sample conditions, the existing methods is challenged by effective information fusion and distinguishable features extraction. To tackle these issues, a few-shot multi-static ISAR target classification method, dubbed multi-level feature reconstruction (MLRF), is proposed in this article, which comprises two key components: inter-station feature reconstruction (ISFR) and inter-class feature reconstruction (ICFR). Specifically, ISFR facilitates complementary information mining and redundant information suppression by multi-static feature alignment and reconstruction, while ICFR enhances inter-class separability and intra-class compactness via multi-class feature interaction and reconstruction. Additionally, a specialized hybrid loss is designed to ensure desired outputs of the related modules. Experimental results on multi-static ISAR dataset of satellite targets demonstrate that the proposed method significantly improves the few-shot classification accuracy, and ablation studies and visual analysis further validate the effectiveness of each module.

Unknown image deformation and few-shot issues have posed significant challenges to inverse synthetic aperture radar (ISAR) target classification. To achieve robust feature representation and precise correlation modeling, this article proposes a novel two-stage few-shot ISAR classification network, dubbed as robust embedding and manifold inference (REMI). In the robust embedding stage, a multihead spatial transformation network (MH-STN) is designed to adjust unknown image deformations from multiple perspectives. Then, the grouped embedding network (GEN) integrates and compresses diverse information by grouped feature extraction, intermediate feature fusion, and global feature embedding. In the manifold inference stage, a masked Gaussian graph attention network (MG-GAT) is devised to capture the irregular manifold of samples in the embedding space. In particular, the node features are described by Gaussian distributions, with interactions guided by the masked attention mechanism. Experimental results on two ISAR datasets demonstrate that REMI significantly improves the performance of few-shot classification and exhibits robustness in various scenarios.

Narrowband and wideband waveforms are usually adopted simultaneously during the observation of micro-motion space targets by inverse synthetic aperture radar (ISAR), which can collect rich multimodal information in the time-Doppler, time-range, and range-instantaneous-Doppler domains. In order to exploit the electromagnetic scattering, shape, structure, and motion characteristics, this article proposes an attention-augmented cross-modal feature fusion recognition network, namely ACM-FR Net. Firstly, the ACM-FR Net adopts convolution neural network (CNN) to extract initial feature vectors from joint time-frequency (JTF) image, high resolution range profiles (HRRPs), and range-instantaneous-Doppler (RID) image, respectively. Then, it transforms the feature vectors of the three modalities into feature sequences. Finally, it achieves interactive feature fusion by implementing attention-augmented cross-modal feature fusion. In the four-category micro-motion space targets recognition experiments, the proposed ACM-FR Net has demonstrated high accuracy and noise robustness.

It is difficult to realize effective synthetic aperture radar (SAR) automatic target recognition (ATR) in open scenarios because the ATR model cannot continuously learn from new classes with limited training samples. When adding new classes to the previously trained model, the capability of recognizing old classes may lose due to severe overfitting. To tackle this problem, a few-shot class-incremental SAR ATR method, namely, hierarchical embedding and incremental evolutionary network (HEIEN), is proposed in this article. First, a hierarchical embedding network and a hybrid distance-based classifier are constructed for basic feature extraction and classification. Then, in order to obtain more accurate decision boundaries, an adaptive class-incremental learning (ACIL) module is designed to adjust the weights of classifiers in all tasks by collecting context information from the past to the present. Finally, a pseudo-incremental training strategy is designed to enable effective model training with only a few samples. Experimental results on the moving and stationary target acquisition and recognition (MSTAR) benchmark data set have illustrated that HEIEN performs well with remarkable advantages in few-shot class-incremental SAR ATR tasks.

For inverse synthetic aperture radar (ISAR) adopting the sparse stepped frequency chirp signal (SSFCS), the target echoes are sparse in the fast time domain with unknown phase errors induced by translational motion, bringing great challenges to well-focused imaging. To tackle this issue, an efficient method for joint motion compensation and high-resolution imaging is proposed under low signal-to-noise ratio (SNR) scenario. Firstly, the signal model is constructed, which is then converted to a probabilistic model with the nonparametric Gamma Process-Complex Gaussian prior. Then, 2D-fast image reconstruction, i.e., 2D-FIR, is proposed for efficient image inference, which avoids matrix inversion by relaxing the lower bound and has high computational efficiency. Finally, a cost function is designed and genetic algorithm is utilized to jointly estimate the translational motion and the 2D image. Experimental results on simulated and measured data have verified the effectiveness of the proposed method.

Restricted by observation conditions, some scarce targets in the synthetic aperture radar (SAR) image only have a few samples, making effective classification a challenging task. Although few-shot SAR target classification methods originated from meta-learning have made great breakthroughs recently, they only focus on object-level (global) feature extraction while ignoring part-level (local) features, resulting in degraded performance in fine-grained classification. To tackle this issue, a novel few-shot fine-grained classification framework, dubbed as HENC, is proposed in this article. In HENC, the hierarchical embedding network (HEN) is designed for the extraction of multi-scale features from both object-level and part-level. In addition, scale-channels are constructed to realize joint inference of multi-scale features. Moreover, it is observed that the existing meta-learning-based method only implicitly utilize the information of multiple base categories to construct the feature space of novel categories, resulting in scattered feature distribution and large deviation during novel center estimation. In view of this, the center calibration algorithm is proposed to explore the center information of base categories and explicitly calibrate the novel centers by dragging them closer to the real ones. Experimental results on two open benchmark datasets demonstrate that the HENC significantly improves the classification accuracy for SAR targets.

In high-resolution radar imaging, the time-varying Doppler induced by maneuvering targets generally invalidates traditional methods. Although sparse signal reconstruction methods can be applied to achieve better performance, the off-grid problem embedded in these methods still prohibits well-focused imaging. In this article, we propose to perform high-resolution radar imaging of maneuvering targets with off-grid scattering centers in a Bayesian framework. First, the statistical model with a parametric dictionary is established, in which the Doppler frequencies of scattering centers are treated as unknown model parameters instead of being discretized into the grid. To be consistent with the physical characteristics, the posterior of the Doppler is approximated by the von Mises distribution. Then, the model parameters and rotation parameters are estimated iteratively by variational inference (VI) and the Newton method. Experiments have demonstrated that the proposed method provides an effective way for high-resolution and well-focused radar imaging of maneuvering targets with off-grid scattering centers in complex scenarios, such as incomplete data and low signal-to-noise ratio.

Scatterer trajectory association is a critical step of 3-D target reconstruction from the inverse synthetic aperture radar (ISAR) image sequence. To cope with the complex scatterer trajectory association of the noncooperative target, a novel method based on multiple hypothesis tracking (MHT) algorithm is proposed. First, the scatterer trajectory association situation is modeled as a multiple hypothesis tree, in which each branch represents a possible association. Then, to generate the hypothesis in each specific branch, a general trajectory motion model is constructed and the parameters are estimated based on the current trajectory association situation. The parameter estimation precision will increase with the growth of the trajectory length. Besides, to eliminate the influence of inaccurate association initialization and direction selection, a fusion algorithm is proposed to merge the forward and backward associated trajectories. Finally, experimental results based on the simulated data and electromagnetic data verify the effectiveness and robustness of the proposed method.

Estimation of the instantaneous attitude of spacecraft plays significant roles in space situation awareness activities, such as on-orbit status monitoring and collision warning. With the rapid development of both optical sensors and inverse synthetic aperture radar (ISAR), it becomes possible to achieve accurate instantaneous attitude estimation of spacecraft by multistatic optical-and-ISAR joint observation. In view of this, this article proposes a novel spacecraft attitude estimation method based on a joint optical-and-ISAR observation system, which includes one optical sensor and two ISARs. Specifically, the proposed method first estimates the orientation and the length of typical components utilizing optical and ISAR images with the same observation instant. Then, it solves the target instantaneous rotation vector from the orientation, length, and Doppler of typical components, and finally, it obtains the instantaneous attitude of the spacecraft. Experimental results have verified the effectiveness of the proposed method.

In complex electromagnetic environments, wideband interference (WBI) may severely affect the imaging quality of synthetic aperture radar (SAR). Because it occupies a large bandwidth, which overlaps with target echoes, the WBI is difficult to mitigate. The existing WBI suppression methods based on filtering or transform-domain analysis usually suffer from a model mismatch. To tackle this problem, a method combining instantaneous frequency (IF) estimation and regularized time-frequency filtering (RTFF) is proposed for WBI suppression and individual components extraction. First, the WBI-corrupted SAR echo is characterized in the time-frequency domain by short-time Fourier transform (STFT) with adaptive window width, determined by the proposed window width optimization method. Then, the IFs of the WBI components are estimated by ridge path detection and regrouping. Finally, the WBI is extracted by RTFF. Experimental results of measured SAR data corrupted by simulated and real WBIs have demonstrated the effectiveness and practicability of the proposed method.

To make full use of the sequential information obtained by continuous inverse synthetic aperture radar (ISAR) imaging, this article proposes a sequential ISAR target classification network based on hybrid transformer (HT). First, a temporal–spatial encoder based on the attention mechanism is designed to extract long-term and global features from sequential images. Meanwhile, a local feature encoder based on the 3-D convolution neural network is designed to extract short-term and local features. Then, the above two features are fused and the classification labels are obtained by a channel encoder–decoder. In 4-satellite target classification experiments, the proposed HT shows high accuracy and robustness to the unknown image scaling, rotation, and combined deformations.

To effectively classify inverse synthetic aperture radar (ISAR) sequential image with unknown deformation, a sequential adjustment ISAR image classification network (SAISAR-Net) is proposed, which first performs global and local image adjustments for each image frame and obtains deformation robust feature sequence. Then, the time-varying features are extracted by attention augmented bidirectional long short-term memory (Bi-LSTM), the output of which is weighted and fused to give a classification label. Compared with the existing deep learning methods, the proposed network significantly improves the classification accuracy and exhibits robustness in scenarios of scaled, rotated, combined transformation, and practical satellite orbit tests.

Restricted by the observation condition, synthetic aperture radar (SAR) automatic target classification based on deep learning usually suffers from insufficient training samples. To tackle this problem, a novel few-shot learning (FSL) framework for SAR target classification, i.e., the mixed loss graph attention network (MGA-Net), is proposed. The classification procedure of the MGA-Net consists of three main stages. In the first stage, the task set is expanded by the data augmentation module to increase diversity. In the second stage, the embedding network is designed to map samples to the embedding space with strong intra-class similarity and inter-class divergence. In the third stage, the multilayer graph attention network (GAT) is constructed and updated according to a novel mixed loss to obtain the classification result. In particular, the data augmentation module alleviates the desire of training samples under large model capacity and enhances the robustness to noise and viewing angle variation; the multilayer GAT accurately captures relations between samples by the attention mechanism; and the mixed loss increases the inter-class separability and accelerates convergence. Experimental results under various few-shot observation settings of the MSTAR and the OpenSARShip benchmark datasets demonstrate that the MGA-Net obtains higher accuracy than typical FSL methods and exhibits robustness to large depression angle variation.

Compared to conventional cameras, the new type of vision camera-neuromorphic cameras, which can avoid motion blur and have the advantages of high spatiotemporal resolution, high dynamic range, low latency, etc. In this paper, two motion compensation methods are introduced for operating event flow based on differential neuromorphic camera, one is motion compensation based on event counting image and time image, another is motion compensation based on image of warped events. Comprehensive experimental study is carried on two datasets, the motion compensation method based on event counting image and time image is better at generating clear motion compensation images, highlighting moving objects inconsistent with the background motion, while the motion compensation method based on image of warped events tends to generate motion compensated images of moving objects with more prominent or clearer edges. Simultaneously, two motion compensation methods are further applied to realize the object detection method based on threshold or maximum contrast. The former method can be adapted in many cases, while the latter method can well detect the translational moving target.

We address the problem of high-resolution radar imaging in low signal-to-noise ratio (SNR) environments in an approximate Bayesian inference framework. First, the probabilistic graphical model is constructed by imposing the sparsity-promoting spike-and-slab prior to the distribution of scattering centers. Then, the model parameters and phase errors are estimated iteratively by expectation propagation (EP) and maximum likelihood (ML) estimation. Compared with the available imaging methods based on the numerical optimization and Bayesian inference, the proposed method has exhibited more flexibility in data representation and better performance in parameter estimation, particularly in sparse-aperture and low SNR scenarios.

This article addresses the problem of joint time–frequency (JTF) analysis of micromotion targets in complex environments in an approximate Bayesian inference framework. First, the sparse observation model is constructed, which is then decomposed into a series of single-window-JTF (SW-JTF) analysis problems to tackle the high dimension of the over-complete dictionary. On this basis, the probabilistic graphical model is constructed by imposing the Gamma-complex Gaussian prior to the JTF distribution. Finally, the model parameters are solved effectively by single-window variational inference (SWVI). Compared with the available methods, the proposed method could obtain better-focused JTF signature for narrowband data and higher quality range-instantaneous Doppler (RID) image for wideband data, especially in low signal-to-noise ratio (SNR) and data corruption scenarios.

Synthetic aperture radar (SAR) automatic target recognition (ATR) plays an important role in SAR image interpretation. However, at least hundreds of training samples are usually required for each target type in the existing SAR ATR algorithms. In this article, a novel few-shot learning framework named hybrid inference network (HIN) is proposed to tackle the problem of SAR target recognition with only a few training samples. The recognition procedure of HIN consists of two main stages. In the first stage, an embedding network is utilized to map the SAR images into an embedding space. In the second stage, a hybrid inference strategy that combines the inductive inference and the transductive inference is adopted to classify the samples in the embedding space. In the inductive inference section, each sample is recognized independently according to a metric based on Euclidean distance. In the transductive inference section, all samples are recognized as a whole according to their manifold structures by label propagation. Finally, in the hybrid inference section, the classification result is obtained by combining the above two inference methods. To train the framework more effectively, a novel loss function named enhanced hybrid loss is proposed to constrain samples to gain better interclass separability in the embedding space. Experimental results on the moving and stationary target acquisition and recognition (MSTAR) benchmark data set illustrate that HIN performs well in few-shot SAR image classification.

Although numerous methods based on sequence image classification have improved the accuracy of synthetic-aperture radar (SAR) automatic target recognition, most of them only concentrate on the fusion of spatial features of multiple images and fail to fully utilize the temporal-varying features. In order to exploit the spatial and temporal features contained in the SAR image sequence simultaneously, this article proposes a sequence SAR target classification method based on the spatial-temporal ensemble convolutional network (STEC-Net). In the STEC-Net, the dilated 3-D convolution is first applied to extract the spatial-temporal features. Then, the features are gradually integrated hierarchically from local to global and represented as the united tensors. Finally, a compact connection is applied to obtain a lightweight classification network. Compared with the available methods, the STEC-Net achieves a higher accuracy (99.93%) in the moving and stationary target acquisition and recognition (MSTAR) data set and exhibits robustness to depression angle, configuration, and version variants.

To obtain high-resolution imaging while reducing the radar operating bandwidth under a low signal-to-noise ratio (SNR), this article proposes a genetic method for accurate residual radial motion estimation and well-focused imaging of the sparse stepped-frequency chirp signal (SSFCS). First, the signal model is constructed by incorporating the residual radial motion parameters into the dictionary. Then, high-quality high-resolution range profiles (HRRPs) are synthesized by Beta process regression (BPR), which has enhanced flexibility in data description and superior performance in parameter estimation. In addition, the genetic method updates the population, i.e., the candidates for the residual radial motion parameters, iteratively according to the image entropy to meet the required precision for well-focused imaging. Experimental results of Monte Carlo simulations and imaging results of simulated and measured data have demonstrated that the proposed method achieves more accurate estimation in residual radial motion parameters and better-focused imaging than the available methods.

Radar echoes simulation has played a significant role in human detection and classification in the scenarios, e.g., antiterrorism, rescue after a disaster and medical, where the real-measured data are generally unavailable and limited. Therefore, a novel radar echoes simulation method of human movements is proposed based on motion capture (MOCAP) and electromagnetic (EM) calculation. First, we generate the trajectories of body segments from the true shape and MOCAP data of a human body. On the basis of that, the radar echoes are simulated by calculating the EM scattering characteristics, i.e., radar cross sections (RCSs) of all the gestures of each body segment's trajectory. Meanwhile, the micro-Doppler characteristics induced by the micromotion of human body segments are modulated in simulated radar echoes. Finally, comparisons between the simulated radar echoes and measured ones prove the validity of the proposed method. Some refinement for RCS calculation of the human body will be investigated in our future work.

This paper addresses the problem of radar-based human gait recognition based on the dual-channel deep convolutional neural network (DC-DCNN). To enrich the limited radar data set of human gaits and provide a benchmark for classifier training, evaluation, and comparison, it proposes an effective method for radar echo generation from the infrared, publicly accessible motion capture (MOCAP) data set. According to the different nonstationary characteristics of micro-Doppler (m-D) for the torso and limbs, it enhances their distinguishable joint time-frequency (JTF) features by applying the short-time Fourier transforms (SFTFs) with varying sliding window length and then designs the DC-DCNN structure to achieve refined human gait recognition by separate feature extraction and fusion. Experiments have shown that compared with the traditional single-channel deep convolutional neural network (SC-DCNN), the proposed method achieves higher recognition accuracy in refined human gait classification without incurring additional radar resources and could be readily extended to refined recognition of other human activities.

We address the problem of image series denoising for high-resolution radar in a nonparametric Bayesian framework. By exploiting the characteristics of amplitude variation at different pixels in the image series, we impose the Gaussian process (GP) model to the corresponding time series of each pixel and achieve effective image series denoising by GP regression. Particularly, the model parameters are solved conveniently by the maximum likelihood estimation. Compared with available denoising techniques in the data domain, spatial domain, and image frequency domain, the proposed method has exhibited more flexibility in data description and better performance in structure preserving and denoising, especially in low signal-to-noise ratio scenarios.

Although the deep convolutional neural network (DCNN) has been successfully applied to target classification of military vehicles based on synthetic aperture radar (SAR), most of the available methods do not fully exploit the characteristics of continuous SAR imaging and only utilize single image for recognition. To extract significant identification features contained in the image sequence, this paper proposes a sequence of SAR target classification method based on bidirectional convolution-recurrent network. In this network, we extract spatial features of each image through DCNNs without the fully connected layer, and then learn sequence features by bidirectional long short-term memory networks. Finally, we design the average softmax classifier to obtain the classification results. Compared with the available methods, the proposed network takes advantage of the significant information in the image sequence and achieves higher classification accuracy in the moving and stationary target acquisition and recognition data set. In addition, it has shown robustness to large depression angle variants, configuration variants, and version variants.

We address the problem of high-resolution radar imaging in complex environments in a Bayesian framework. We perform model order selection and sparse weights estimation via the maximum a posterior-expectation maximization technique in a statistical model, where the noise obeys Gaussian mixture distribution; and the weights are governed by the sparsity-promoting Gamma-Gaussian hierarchical prior. The proposed method has closed-form solution and can be implemented efficiently by matrix operation. Experiments has shown that accounting for the noise with Gaussian mixture model instead of single Gaussian greatly improves the performance, and the proposed method provides an effective way of high-resolution radar imaging in complex environments such as barrage jamming and micro-Doppler interference.

Although the deep convolutional neural network (DCNN) has been successfully applied to automatic target recognition (ATR) of ground vehicles based on synthetic aperture radar (SAR), most of the available techniques are not suitable for inverse synthetic aperture radar (ISAR) because they cannot tackle the inherent unknown deformation (e.g., translation, scaling, and rotation) among the training and test samples. To achieve robust polarimetric-ISAR (Pol-ISAR) ATR, this paper proposes the spatial transformer-multi-channel-deep convolutional neural network, i.e., ST-MC-DCNN. In this structure, we adopt the double-layer spatial transformer network (STN) module to adjust the image deformation of each polarimetric channel and then perform a robust hierarchical feature extraction by MC-DCNN. Finally, we carry out feature fusion in the concatenation layer and output the recognition result by the softmax classifier. The proposed network is end-to-end trainable and could learn the optimal deformation parameters automatically from training samples. For the fully Pol-ISAR image database generated from electromagnetic (EM) echoes of four satellites, the proposed structure achieves higher recognition accuracy than traditional DCNN and MC-DCNN. Additionally, it has shown robustness to image scaling, rotation, and combined deformation.

Space debris damage orbiting spacecraft and astronauts and ISAR imaging is an important method to recognize and classify debris. Compared with 2-D imaging, 3-D imaging is able to provide more information. However, debris with rapid spinning have great migration through range-cells, so common methods are unproductive. A novel method of ISAR 3-D imaging based on nonparametric Bayesian model is proposed aimed at debris with spinning. Firstly, a motion model and a signal model are proposed. Secondly, PSO algorithm is utilized to preprocess the data and obtain the height of target. Finally, Nonparametric Bayesian model is imposed to elaborately reconstruct the target in range and cross-range. For monostatic radar, point-target simulation data and electromagnetism data confirm that the method will obtain refined 3-D imaging results. Meanwhile, this method is capable to surmount the obstacle of Doppler aliasing and data missing caused by rapid spinning.

In recent studies, synthetic aperture radar (SAR) automatic target recognition (ATR) algorithms based on convolutional neural network (CNN) have achieved high recognition rates in the moving and stationary target acquisition and recognition (MSTAR) data set. However, the correlation between clutter in the training and test data sets is ignored in these algorithms, although most of them used only the center part of the images by removing a lot of the clutter but not everything, which may result in better performance than what would be achieved in the operational scenarios. To tackle this problem, we propose a target segmentation method based on morphological operations to generate data sets without clutter. Then, we design the large-margin softmax (LM-softmax) batch-normalization CNN (LM-BN-CNN) structure, which utilizes the LM-softmax classifier in the last layer to increase the separability of samples after clutter removal. In addition, this structure performs BN with constant mean and variance to increase the convergence speed and reduce overfitting. Experiments on the MSTAR data set have shown that LM-BN-CNN obtains better performance than the available CNNs designed for SAR ATR of ground vehicles, and it has robustness to large depression angle variation, configuration variants, and version variants.

The anisotropy of radar cross section of scatterers makes the scatterer trajectory matrix incomplete in sequential inverse synthetic aperture radar images. As a result, factorization methods cannot be directly applied to reconstruct the 3-D geometry of scatterers without additional consideration. We propose a modified expectation-maximization (EM) algorithm to retrieve the complete scatterer trajectory matrix. First, we derive the motion dynamics of the projected scatterer, which approximates an ellipse. Then, based on the estimated ellipse parameters using the known data of each scatterer trajectory, we use the Kalman filter to initialize the missing data. To address the limitations of a traditional EM, which only considers the rank-deficient characteristics of the scatterer trajectory matrix, we propose to augment EM by using both the known rank-deficient and elliptical motion characteristics. Experimental results on simulated data verify the effectiveness of the proposed method.

To obtain higher range resolution without incurring significant hardware costs, this paper proposes a novel method for high-resolution sparse subband imaging based on Bayesian learning. The signal model is derived and a probabilistic model is constructed. In particular, hierarchical sparse-promoting priors are imposed on the distribution of scattering centers, which is conjugate to the likelihood function. Then, a closed-form solution is derived based on the MAP-expectation-maximization framework. A multilevel dictionary which automatically adjusts the distance between adjacent atoms is adopted to achieve refined estimation with moderate computational burden. Finally, a coherent processing method is addressed. Experimental results have demonstrated the effectiveness of the proposed method in low signal-to-noise ratio and complex target scenarios.

To realize nonparametric imaging of targets in complex micromotion, the available method achieves scattering center trajectory association by Kalman filtering with the minimum Euclidean distance criterion. Then, it solves the 3-D distribution of scattering centers using matrix factorization. For targets with complex structure, however, the Kalman filter usually leads to wrong association due to serious trajectory intersection in the range-slow time domain. To tackle this problem, this paper proposes an effective trajectory association method based on the 2-D range-instantaneous-Doppler image series. Additionally, it refines the trajectory matrix using modern spectral analysis to facilitate accurate 3-D imaging. Simulation results have proved the effectiveness of the proposed method.

This paper presents an automatic ship detection algorithm for polarimetric synthetic aperture radar (PolSAR) data. Based on the non-Gaussian K-Wishart distribution model for complex backscattering coefficients, the PolSAR image is clustered automatically by a modified expectation maximization algorithm. A goodness-of-fit test is incorporated to improve the model fitness of the cluster iteratively. Then, the SPAN of ship cluster center is used to detect ships. Finally, the experimental results of a real measured UAVSAR dataset show that the proposed algorithm could improve the ability of weak target detection while reduces the rate of false alarm and miss detections.

Traditionally, the factorization method is applied to reconstruct the 3D geometry of a target from its sequential inverse synthetic aperture radar images. However, this method requires performing cross-range scaling to all the sub-images and thus has a large computational burden. To tackle this problem, this paper proposes a novel method for joint cross-range scaling and 3D geometry reconstruction of steadily moving targets. In this method, we model the equivalent rotational angular velocity (RAV) by a linear polynomial with time, and set its coefficients randomly to perform sub-image cross-range scaling. Then, we generate the initial trajectory matrix of the scattering centers, and solve the 3D geometry and projection vectors by the factorization method with relaxed constraints. After that, the coefficients of the polynomial are estimated from the projection vectors to obtain the RAV. Finally, the trajectory matrix is re-scaled using the estimated rotational angle, and accurate 3D geometry is reconstructed. The two major steps, i.e., the cross-range scaling and the factorization, are performed repeatedly to achieve precise 3D geometry reconstruction. Simulation results have proved the effectiveness and robustness of the proposed method.

This paper presents recent advances in high-resolution radar imaging and parameter estimation of aerospace targets with micromotion, which assume that the radar operates in the high frequency regime and the point scattering model holds. Particularly, principles of the parametric and nonparametric imaging methods are introduced, discussed, and demonstrated using simulated and measured data.

When wide or ultrawideband, low pulse repetition frequency (PRF) radar is applied to the imaging of space targets; it is highly possible that motion through range cell and azimuth under-sampling occurs, which will result in image smearing. To figure out this problem, this paper proposes a novel, three-step imaging method using the high-resolution range profile (HRRP) series. In the first step, high-quality HRRP series are obtained based on the theory of sparse signal representation. Then, based on the Kalman predictor and the minimum Euclidean distance criterion, motion and amplitude feature-based scatterer trajectory association is carried out to form the scatterer trajectory matrix, from which the scatterer locations are conveniently solved in the last step. Compared to the traditional imaging techniques based on Doppler analysis, the proposed method is able to mitigate the influence of azimuth under-sampling, and may provide a new solution to high-resolution imaging of targets moving nonuniformly in low PRF scenarios. Finally, simulations have proved the effectiveness of the proposed method.

To achieve finer range resolution without increasing the bandwidth and sampling rate of the radar system, high-resolution imaging by data synthesizing using sparse subbands has received intensive attention in recent years. This paper derives the imaging geometry and signal model for radar imaging of space targets from sparse subbands. Next, it introduces and analyzes the available methods. Then, a practical method is proposed for sparse subband imaging of space targets in high-speed motion, which comprises phase compensation along the range and azimuth, gapped-data filling, scatterer number estimation, amplitude estimation, and azimuth image fusion. Finally, imaging of the simulated data using the fixed-point and electromagnetic scattering models proved the validity of the proposed method.

The narrow-band interference (NBI) is a common jamming signal against synthetic aperture radar (SAR), which can degrade the imaging quality severely. This paper proposes a new method for NBI suppression in the data domain based on the independent component analysis (ICA). In this method, echoes contaminated by the NBI are identified in the frequency domain. Next, time filtering and whitening are performed to the identified echoes. Then, the ICA is carried out to decompose the echoes into a series of basis signals, and the jamming components are selected by thresholding. Finally, the NBI is reconstructed and subtracted from the echoes, and the well-focused SAR imagery is obtained by conventional imaging methods. The proposed method copes well with the time-varying NBI with little signal loss. Results of simulated and measured data have proved the validity of the proposed method.

Based on the synthetic aperture radar (SAR) geometric model, a novel, fast algorithm of large scene deceptive jamming against the space-borne SAR is proposed. First, we divide the jamming scene template into sub-templates according to the depth of focus in the range dimension. Next, each sub-template is decomposed into the slow-time-dependent and slow-time-independent terms in the range frequency-azimuth time domain. The slow-time-independent terms are generated off-line while the slow-time-dependent terms are generated by real-time 1-D frequency modulation. Then, the sub-templates are convolved with the intercepted SAR signals simultaneously. Finally, fast deceptive jamming is achieved by incorporating all the sub-templates together. In the proposed method, the two-step realization of the sub-templates and the parallel sub-block processing improves the algorithm efficiency. The simulation results prove the validity of the proposed algorithm.

Based on the difference in statistical distribution between the target and jamming signal in the synthetic aperture radar (SAR) image, this letter proposes a novel adaptive method for barrage jamming suppression. In this method, the covariance matrix of jamming is estimated from the SAR image. Then, the 2-D sinc function of the ideal point target is utilized as the steering vector to obtain the optimal adaptive filter. This filter can suppress the random barrage jamming effectively, thus improving the image contrast and interpretability. Additionally, this letter analyzes in detail the theoretical basis and performance of the proposed method. Finally, simulations are provided to demonstrate its effectiveness.

This paper proposes a novel method for high-resolution imaging of group targets moving in a formation with constant accelerated rectilinear motion. In this method, the first- and second-order phase terms are compensated simultaneously to obtain a “bulk” image of group targets. Then, regions of subtargets are determined by the segmentation method based on clustering number estimation and normalized cuts. Finally, refined compensation of the second- and third-order phase terms is carried out jointly to obtain a well-focused image of group targets, following the minimum local image entropy criterion. Simulation results have proved the validity of the proposed method.

In 3-D inverse synthetic aperture radar (ISAR) imaging of spinning space debris, the image obtained via the available algorithm is modified by a scaling factor. Determined by the angle between the spinning vector and the radar line of sight, this factor cannot be estimated by a monostatic radar in a short imaging interval. This letter derives the bistatic ISAR (Bi-ISAR) geometry and signal model for 3-D imaging of space debris. Then, a 3-D image scaling algorithm is introduced based on the connections between the mono- and bistatic echoes of the same scatterer. Numeric simulations have proved the validity of the proposed algorithm.

Narrow-band interference (NBI) is a common interference source in synthetic aperture radar (SAR) imaging. Its existence will degrade the imaging quality greatly. Based on detailed analysis on the characteristics of NBI, this letter proposes a new NBI suppression algorithm using the complex empirical mode decomposition (CEMD) method. In this algorithm, echoes that include NBI are recognized in the time domain first. Then, these echoes are decomposed into a number of intrinsic mode functions (IMFs) via the CEMD. After that, IMFs that correspond to NBI are subtracted from the echoes by thresholding. Finally, well-focused SAR imagery can be obtained from the separated target echoes using traditional SAR imaging algorithms. The effective data loss in this algorithm is smaller than other NBI suppression approaches. In addition, this algorithm is robust to time-varying NBI. Imaging results of measured data have proved the validity of this algorithm.

Since space debris could post a significant threat to orbiting objects around the Earth, their reorganization, measurement, and catalogue are of great importance. This paper establishes a 3-D inverse synthetic aperture radar (ISAR) imaging geometry and signal model for space debris. Then, a 3-D imaging algorithm is proposed to realize coherent imaging in the range-slow-time domain. This algorithm is based on the complex-valued back-projection transform according to the spinning nature of space debris. The simulation results for both point scattering and continuous targets have proved the validity of the proposed algorithm.

For micromotion targets with rotating parts, the inverse synthetic-aperture-radar image of the main body may be shadowed by the micro-Doppler. To solve this problem, this paper proposes an imaging algorithm based on the complex-valued empirical-mode decomposition. First, the radar echoes are decomposed into a series of complex-valued intrinsic-mode functions (IMFs). Then, the IMFs from the rotating parts and those from the main body are separated according to the characteristics of their zero-crossings. Finally, the well-focused imaging of the main body via traditional imaging algorithm and the accurate parameter estimation of the rotating part can be obtained. Both the imaging results for the simulated and measured data are given to verify the validity of the proposed algorithm.