Radar transmits active agile waveformplays a significant role in anti-jamming. However, efficient coherent integration of target's agile echo in a coherent processing interval (CPI) usually poses a severe challenge. This letter proposes an improved sparsity adaptive matching pursue (ISAMP) to address this issue. Firstly, the agile echo signal model of random interpulse frequency and PRT joint agile (RI-FPrtJA) waveform is derived; The sparse reconstruction model of RI-FPrtJA echo signal is then mathematically deduced based on compress sensing theory; Lastly, the ISAMP is proposed to accurately accomplish sparse reconstruction based on the regularized grids mismatch correction and adaptive searching step extension. The simulation results demonstrate that the ISAMP method can achieve better coherent integration in terms of mismatch sidelode levels suppression, sparse reconstruction capacity, and computational cost, compared to some current existing methods.

Remote sensing scene classification categorizes imaged areas by ground contents to provide semantic information for Land Use and Land Cover monitoring. The remote sensing images exhibit diverse scene targets, textures, shapes, and spatial arrangements under varying solar illumination, therefore needing high-level feature analysis for effective scene characterization. Existing methods, including the widely used conventional neural networks (CNNs), often fall into local minima during the optimization process due to the non-convexity of the loss function over the parameter space. Furthermore, existing methods have two limitations in feature extraction: single-scale features are not adaptive to the spatial variation, and simply concatenating multi-scale features neglects the geometric relationships among resolutions. Moreover, the flattening operation commonly used in network architectures tends to destroy the inherent topological structure of images, leading to a loss of spatial information in the scene representation. To address these issues, in this study, a Chaos-based multi-scale tensor pyramid network (CMTPN) is proposed. Firstly, a 3D discrete chaotic system is proposed based on both the logistic map and the tent map to achieve a more uniform distribution and a broader full chaotic map range. The map is then embedded into the neural network structure with the aim of obtaining an optimization result. Secondly, a multi-scale tensor pyramid network is proposed. This network extracts local geometric details and global contextual information through parallel branches, in which the Tensor Contraction Layer is utilized to combine features from different scales by using attention mechanisms to adaptively integrate high- and low-resolution information, and meanwhile the Parallel Tensor Regression Layer breaks down the multi-scale features using the Tucker decomposition for the final classifications. Effectiveness of the proposed algorithm is demonstrated through extensive experiments conducted on three remote sensing benchmark datasets: NWPU-RESISC45, AID, and OPTIMAL-31, showing that the proposed CMTPN not only achieves superior classification accuracy but also maintains competitive computational efficiency, robustly outperforming state-of-the-art methods.

This work introduces DiSpeckle, a diffusion-based SAR despeckling method in which a physics-inspired stochastic differential equation (SDE) governs the bidirectional transition process between speckled and clean images. The key insight is that the forward diffusion process closely parallels the physical formation of speckle noise—both involve the accumulation of microscopic random perturbations that progressively distort the underlying signal. Building on this analogy, the forward diffusion process is designed using an SDE that emulates the speckle formation mechanism, namely the coherent aggregation of unresolved backscattered signals. Despeckling is then achieved by inverting this process through the corresponding reverse SDE. Unlike previous diffusion-based methods that rely on transforming speckle into an approximately Gaussian distribution, this physics-inspired formulation offers a mathematically rigorous and physically consistent pathway from speckled to clean imagery. In addition, DiSpeckle provides a flexible framework that seamlessly integrates off-the-shelf Gaussian denoisers. Experiments demonstrate that DiSpeckle—even when using off-the-shelf Gaussian denoisers—matches or outperforms state-of-the-art despeckling methods. With fine-tuned denoisers, it achieves superior performance while requiring only one-third of the training data. The code will be made publicly available on GitHub upon publication.

This paper investigates the effect of current-wave interaction on the ocean surface current (OSC) retrieval under the DCA framework using an improved Doppler radar imaging model (IDopRIM). Both numerical simulation data and a real SAR observation image in scenarios involving ocean internal waves (IW) are utilized for analyses. Experimental results suggest that the contribution of current-wave interaction to the sea surface Doppler velocity cannot be ignored for the IW ocean surface with rapidly varying currents. Under certain conditions, the relative errors of the OSC retrieval due to neglecting such a contribution could be larger than 30%.

Polarimetric High-Resolution Range Profile (HRRP) holds significant potential in Radar Automatic Target Recognition (RATR) due to its ability to provide detailed polarimetric and spatial information. In recent years, deep learning methods have been extensively applied in RATR based on polarimetric HRRP. However, these methods often focus solely on local or temporal features, thereby not fully utilizing the spatial information. Additionally, valuable polarimetric information is frequently overlooked. Moreover, most of these methods do not distinguish between target and noise areas, resulting in a lack of emphasis on crucial range cells. This paper introduces a method for polarimetric HRRP recognition based on the Vision Transformer (ViT), which effectively extracts both local and temporal features. Our approach incorporates a polarimetric preprocessing step in which manual features and Convolutional Neural Networks (CNNs) are combined to enhance the extraction of polarimetric features. To direct the network’s focus toward significant range cells, we design a novel attention loss. Experimental results demonstrate that our proposed method improves recognition accuracy and maintains robustness in noisy environments.

The active jamming suppression of Synthetic Aperture Radar (SAR) images remains a severe challenge. The jamming SAR images simulation dataset is firstly built by open SSDD dataset and random shift-frequency jamming type; The formula of existing diffusion model is then modified using the low-rank based block space filter (BSF) theory; Lastly, jamming SAR images are as the inputs to effectively train our proposed model to generate the jamming suppressed images. Experimental results qualitatively and quantitatively demonstrate the effectiveness of the proposed method.

In this paper, we propose an edge attention superpixel segmentation method for polarimetric synthetic aperture radar (PolSAR) images. Image edges are specially considered during the segmentation process by imposing an edge-attention-based distance. The edge-attention-based distance is defined on an edge map and it ensure that one superpixel is not divided into several parts by image edges, which helps to generate superpixels with better boundary adherence. Experimental results on AirSAR and Radarsat-2 images demonstrate the superiority of the proposed method.

SAR despeckling is a crucial preprocessing step for various SAR applications. This work introduces SpeckleDiff2D, a complex-valued speckle reduction diffusion network. Our approach emphasizes the design of a forward noise-accumulation process that mimics the generation mechanism of speckle noise. The proposed despeckling algorithm is trained exclusively on synthesized speckled images generated by optical dataset and is validated on both simulated and real SAR images. Experimental results demonstrate the method’s efficiency and resilience to domain shifts, positioning it as a promising solution to the persistent challenge of data insufficiency in SAR despeckling design.

In this paper, the characterization of different target scattering vector models for the physical scattering mechanism of a target is compared, focusing on the differences between our proposed target scattering vector model and other classical scattering vector models. The comparison will be in two parts, depending on the type of data in use. For the analysis of the physical scattering mechanism of single-look data, the α/β scattering vector model, the con-diagonalization scattering vector model developed by Touzi, and our proposed αBscattering vector model are compared for the effect of characterizing the physical scattering mechanism of the target, respectively. For the analysis of the physical scattering mechanism of multi-look data, the effects of H/α decomposition and our proposed ∆αB/αBdecomposition on the classification of the target areas are compared, and the results are combined with the Wishart iterative classifier to optimize the classification results further. The experiments are performed on RADARSAR-2 (C-band) and ALOS-2 (L-band) data from San Francisco, USA. The results show that for single-look data, the αBscattering vector model is superior in overall performance. For multi-look data, the H/α and ∆αB/αBdecompositions tend to misclassify volume and double scattering, respectively, and the ∆αB/αBmay have better results than the H/α decomposition for the application of man-made targets dominated by double scattering.

Using polarimetric synthetic aperture radar (PolSAR) imagery for ship detection is a critical research area in marine surveillance. Currently, the mainstream methods primarily fall into two categories: superpixel approaches and neighborhood matrix methods. These methods aim to utilize both the polarimetric and spatial information of the neighborhood pixel patch for detection. However, existing methods may not fully exploit the potential of neighborhood information. This letter formulates the ship detection problem as a binary classification task and introduces an innovative ship detection algorithm based on kernelized support tensor machine (K-STM). By employing neighborhood polarimetric tensors as the feature representation of the pixel patch, we can implicitly incorporate all polarimetric and spatial information within different dimensions of the tensor. With the help of the tensor kernel function, K-STM can effectively extract feature information embedded in the neighborhood polarimetric tensors across different dimensions. Two PolSAR datasets acquired from Radarsat-2 are used for experimental validation. The proposed K-STM method achieves the highest figure of merit (FoM) of 0.898 and 0.975 for two datasets. It demonstrates that the proposed method can achieve better performance on ship detection.

Local climate zone (LCZ) classification plays a critical role in urban environment research and has attracted extensive attention from many researchers. However, the potential of deep learning-based approaches is not yet fully explored in this field, even though neural networks continue to push the frontier for various applications. In this paper, we propose a novel multimodal multiscale Transformer network for LCZ classification by introducing multiscale patch embedding and multimodal fusion learning in Transformer architecture. The proposed multiscale patch embedding effectively captures hierarchical interrelationships of image contextual neighborhoods, and automatically learns discriminative features. And the proposed multimodal fusion learning enables the network to naturally fuse multispectral and synthetic aperture radar (SAR) data under the guidance of attention mechanism. To further improve classification accuracy, we impose semi-supervised learning to mine unlabeled image data information. Both labeled and pseudo-labeled data jointly drive our network updates. Experiments conducted on the So2Sat LCZ42, CHN15-LCZ and SouthKorea6-LCZ benchmark datasets demonstrate that our proposed approach outperforms other existing methods significantly and achieves state-of-the-art performance. In the generated LCZ maps, urban and natural classes are well distinguished, the urban structure with waters or mountains is well preserved. Finally, we also discuss the impact of the sample receptive field and sample heterogeneity on LCZ classification performance, which provides a new idea for future studies of LCZ classification.

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.

This paper studies and summarizes a feature extraction strategy from the multiscale monogenic space for target recognition in synthetic aperture radar (SAR) images. We propose applying the regional Euclidean (L2) norm to extract the features of the monogenic signal. It can obtain the maximum eigenvalue in each region and reduce information loss. We performed several groups of experiments on dimensionality reduction, they show that the regional L2 norm extraction method has the best classification results. MSTAR, OpenSARship, and FUSAR-ship datasets are the commonly used verification datasets for SAR target recognition. The experiments on these datasets have demonstrated the effectiveness of the monogenic signal extracted with regional L2 norm for SAR target recognition. In combination with K-Nearest Neighbor (KNN), sparse representation classifier (SRC), or support vector machines (SVM), the proposed method achieves great results compared with some advanced algorithms.

Multi-band polarimetric synthetic aperture radar (PolSAR) has significant advantage in information extraction. However, the demanding acquisition requirement greatly prohibits its development. Typically, compared to low-frequency band PolSAR data, high-frequency band suffers more severe data insufficiency. In this paper, the authors proposed to resolve this issue by simulating Ka-band PolSAR images from X-band images. For this purpose, a conditional Generative Adversial Network (cGAN) based X-to-Ka band PolSAR image transfer network has been proposed. Adaptations in terms of preprocessing and loss function are made to the original cGAN so that it can be better adapted to PolSAR image processing. The proposed method is verified using the X- and Ka-band dataset acquired in Hainan, China by the Aerial Remote Sensing System of the Chinese Academy of Sciences. Experimental results demonstrate the feasibility of the proposed method.

With the change in the spatial angle, the cross-polarization isolation (CPI) of polarimetric phased-array radar (PPAR) changes as well, destroying the estimation of the target polarization scattering matrix (PSM). To correct the bias in PPAR, this article comprehensively designs the transmitting antenna, receiving antenna, and signal waveform and proposes a bias correction method based on system optimization. First, for the transmitting antenna of PPAR, the second-order cone program (SOCP) model is proposed to optimize the weighting coefficient. With the SOCP-based beamforming method, not only beam pattern in any spatial angle can be achieved, but also arbitrary polarization state can be precisely configured. Then, for the wideband receiving signal with a certain beamwidth, an angle estimation method based on eigenvalue decomposition is proposed in this article, which can effectively cure the challenges introduced by the beamwidth and signal bandwidth. Subsequently, for the signal waveform, the phase code is designed to measure all the elements of PSM in simultaneous transmission and simultaneous reception (STSR) mode, which could eliminate the biases of the moving speed and the second-order cross-polarization error. Finally, this article compares with other methods based on differential reflectivity, and experiments show that the factors such as the spatial angle, array structure, antenna beamwidth, signal bandwidth, motion speed, and signal-to-noise ratio (SNR) have the least influence on the present method in this article.

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.

Data insufficiency poses a significant challenge in Ka-band Polarimetric Synthetic Aperture Radar (PolSAR) applications. Traditional PolSAR simulation approaches fail to conquer this issue due to the intricate modeling and computational complexities induced by high-frequency. In this paper, the authors propose to mitigate this issue through neural style transfer. An X2Ka translation network is proposed to transfer X-band PolSAR images to Ka-band. Leveraging the well-verified generative network Pix2Pix, the authors adapt it to accommodate the specific discrepancies between PolSAR and optical data. Experiments are conducted on X- and Ka-band PolSAR images acquired by an Airborne PolSAR system from the Chinese Academy of Sciences. Both qualitative and quantitative evaluation results demonstrate the effectiveness of the proposed network.

This paper numerically investigates the spatial ergodicity of radar scattering from randomly rough ocean surface. Based on the accurate full-wave multi-level steepest decent-sparse matrix canonical grid (MLSD-SMCG) method and Monte Carlo simulation, we simulate the L-band normalized radar scattering coefficients from one-dimensional (I-D) rough ocean surfaces with different radar illumination sizes. According to the simulation results, it is found that: For the scattering angle less than 85°, the normalized bistatic radar scattering coefficient from ocean surface with radar illumination size no less than $16\lambda$ has good spatial ergodicity. Also, the emissivities from ocean surfaces with sizes exceeding $64\lambda$ manifest spatial ergodicity if the measurement accuracy of emissivity is less than the order of magnitude of 10 −4 .

In this study, the eigenvector-based polarimetric feature extraction methods are analyzed based on the target dichotomy decomposition. The eigenvector-based scattering characterization methods, including the α / ß scattering model, Touzi's scattering model, and a recently developed scattering vector parameterization model, are applied to the single scattering matrix decomposed by the stable Huynen decomposition, i.e., Yang's decomposition. Experiments are performed on RADARSAT-2 data acquired over an oil spill area, where several scattering behaviors are presented such as backscatter from ships and the metallic target, ocean surface, oil slicks, and the multiple reflections between ships and ocean surface. Results show that the dichotomy decomposition-based scattering vector parameters are effective in characterizing the dominant scattering mechanism of targets.

Change detection is an important topic for the use of polarimetric synthetic aperture radar (PolSAR) images. In this study, we investigate the change detection method based on regions. First we segment PolSAR images with improved simple linear iterative clustering (SLIC), then take the averaged coherency matrix as the following change detection unit. The performances of three classical test statistics based on regions and pixels are compared, respectively. Three data sets collected over Langqi Island, Fuzhou, China, by RADARSAT-2 are used for evaluation. Experiments show the region-based method can stably improve the performance of change detection and reduce the false alarms than those of the pixel-based method. Further, the optimization of polarimetric contrast enhancement (OPCE) test statistic is not easily affected by spackle noise and provides stable detection results.

In this letter, we present a novel ship detection method for polarimetric synthetic aperture radar (PolSAR) images. Generalized polarization relative entropy (GPRE) is proposed to measure the differences between the target and clutter in scattering mechanism, randomness, and intensity. Since it is difficult to derive a theoretical closed-form of the GPRE, we employ the kernel density estimation to model the distribution of the GPRE in ocean regions. Then, a constant false alarm rate (CFAR) ship detection method is proposed based on the estimated distribution. Experiments performed on both synthetic and real scene images demonstrate the effectiveness of the proposed method.

This letter proposes a novel convolutional neural network (CNN) method for dual-polarized synthetic aperture radar (SAR) ship grained classification. The network employs hybrid channel feature loss that jointly utilizes the information contained in the polarized channels (VV and VH). It is demonstrated that, by adopting the proposed CNN framework and the novel loss function, the classification performance can be efficiently improved. First, instead of the prevalently used threefold or fourfold division (container ship, oil tanker, bulk carrier, and so on), the proposed method can further divide vessels into eight accurate categories. Second, this method can not only effectively classify targets into eight categories but also its accuracy in terms of fewer category classifications surpasses existing methods. Third, the method can achieve good performance on a small training data set. Experiments conducted on the OpenSARShip data sets indicate that the proposed classification method achieves state-of-the-art results.

Synthetic aperture radar (SAR) plays an essential role in earth observation and projection due to its capability to penetrate clouds, which makes it possible to monitor terrestrial surfaces under all weather conditions. Multiplicative noise often occurs in the SAR signal, hampering the retrieval of information from SAR imagery. Convolutional neural networks (CNNs) have been used in many computer vision tasks and are helpful in image denoising. However, current CNN-based denoising approaches inevitably lead to a “washed out” effect that loses spatial details. Another limitation is that most typical CNN-based denoising models require a noise-free image for training. To address these issues, we propose a novel end-to-end self-supervised SAR denoising model: Enhanced Noise2Noise (EN2N), which can be trained without a noise-free image. To enhance the quality of the result images, the perceptual features from a pre-learned CNN are introduced to restore the spatial details by a hybrid loss function. Experiments show that our proposed method outperforms the typical denoising methods in terms of noise reduction and feature preservation based on image quality metrics. Also, the new hybrid loss could enhance the spatial details significantly. The good performance maintains the robustness throughout time, which reduces the uncertainty in time-series SAR caused by random noise. Benefiting from optimization of graphics processing unit (GPU) and multi-threading, the proposed method has higher computation efficiency than traditional methods. This study demonstrates the great potential of using our self-supervised deep learning approaches for SAR image denoising in the future.

Constant false alarm rate (CFAR) detector is a class of widely used methods for target detection in radar images. Classical CFAR detectors perform target detection on a pixel-by-pixel basis using certain sliding windows for estimating clutter statistics, which run fast for small images. However, as the image size gets large, the time cost of these detectors will increase significantly since the time complexity with respect toN×N-pixel image isO(N2). In practice, radar images, such as those in synthetic aperture radar (SAR), usually have very large numbers of pixels (which can be on the order of 10000 × 10000), making the classical CFAR detectors very time-consuming when applied to these images. In this paper, we present graphics processing unit (GPU)-oriented Designs for speeding up CFAR detectors, including smallest/greatest-of CFAR and order-statistic CFAR. The proposed designs implement CFAR detectors via tensor operations, including tensor convolution, shift, and boolean operation, which can be fast operated by GPU. Experiment results show that the proposed GPU-oriented CFAR detectors running on a high-performance Nvidia RTX 3090 GPU can be thousands of times faster than the classical CFAR detectors, and realize real-time target detection in large-size radar images. Examples using SAR and range-Doppler images are provided as illustrative applications of the proposed GPU CFAR detectors to target detection in radar images.

Radio spectrum bands allocated to spaceborne synthetic aperture radar (SAR) imagery are shared by multiple missions. In practical radio spectrum environments, these bands are also used by some ground radars, e.g., C-band weather radar. Due to this fact, radio frequency interference (RFI) may occur for a spaceborne SAR when its received signals contain the transmitted waveforms from another SAR or radar operating at the same frequency band. This particular class of RFI is usually linear-frequency-modulation (LFM) signals, which can cause bright radiometric artifacts in focused SAR images. Most existing signal processing approaches designed for addressing this problem belong to the class of preprocessing methods, which removes RFI in level-0 raw radar data before SAR focusing. In this article, we propose a postprocessing kernel—2-D SPECtral ANalysis (2-D SPECAN) filter, for removing the class of LFM RFI in level-1 SLC images. The filtering consists of three main steps: Step 1: focus LFM RFI artifacts in SLC images as point-like responses in the spectral domain via 2-D SPECAN; Step 2: perform 2-D notch filtering in the spectral domain to remove the most contribution of the RFI responses; and Step 3: transform the filtered spectrum back into the SLC image domain using the inverse operation of the 2-D SPECAN. For computation efficiency, we design a simplified processing flow and adopt a blockwise processing strategy. Experiments with several Sentinel-1 SLC images demonstrate that severe RFI artifacts in SLC images can be removed significantly by the proposed method.

Superpixel segmentation approaches for polarimetric synthetic aperture radar (SAR) images have only been studied in recent years. Simple linear iterative clustering (SLIC) is a simple and efficient superpixel segmentation method, first proposed for optical images. It basically includes three implementation steps, i.e., initialization, local$k$-means clustering, and postprocessing. The challenge of applying SLIC to polarimetric SAR images lies in constructing the effective spatial and feature similarity and proposing the efficient segmentation procedure. In this study, to address both issues, we modify the SLIC clustering function to adapt the characteristics of polarimetric statistical measures. A new initialization method is proposed, which exploits the image gradient information to produce robust cluster centers. Furthermore, in an effort to give a comprehensive comparison and provide a fair assessment of the feature similarities for polarimetric SAR imagery, four classic statistical distances, among which two were not studied along with the SLIC previously, are embedded in the modified clustering function. The proposed method is validated by comparing with state-of-the-art SLIC-based algorithms and also the Ncut and TurboPixel algorithms. Experiments on extensive polarimetric SAR data sets show that the proposed method can significantly improve the segmentation results, with producing better boundary adherence and compact as well as uniform superpixels. We also obtain distinct conclusions that are different from the existing studies when investigating the performances of the statistical measures.

Pseudo quad polarimetric (quad-pol) image reconstruction from the hybrid dual-pol (or compact polarimetric (CP)) synthetic aperture radar (SAR) imagery is an important technique for radar polarimetric applications. There are three key aspects concerned in the literature for the reconstruction methods, i.e., the scattering symmetric assumption, the reconstruction model, and the solving approach of the unknowns. Since CP measurements depend on the CP mode configurations, different reconstruction procedures were designed when the transmit wave varies, which means the reconstruction procedures were not unified. In this study, we propose a unified reconstruction framework for the general CP mode based on one of our previous work. Both the iterative and least squares (LS) error solving approaches are extended. Validation is carried out on polarimetric data sets from RADARSAT-2 (C-band) to compare the performances of the reconstruction models and CP modes.

To solve the problem of dense vehicle target detection in polarimetric synthetic aperture radar (PolSAR) images from urban areas under complex scenarios, this paper proposes a target detection method that combines the superpixel segmentation and the Wishart classifier. Firstly, the buildings are detected based on the different polarimetric scattering characteristics of ground objects. Then, the morphological information of the target is obtained by the local Wishart classifier and the superpixel segmentation. After that, the center points of the target are obtained by the global Wishart classifier. Finally, the region growing procedure is used to fuse the information obtained by above-mentioned classifiers to complete the target detection task.

In this letter, we first investigate the relationship between polarimetric covariance matrix [C] and complete polarimetric covariance difference matrix [CP], and then construct a scattering difference parameter SDP. Subsequently, a hybrid polarimetric covariance matrix [HC] is developed based on SDP for curing the disadvantage of [CP], that is the scattering difference information of small ships cannot be well contained in [CP]. By fusing the feature “1-SDP” and the power detector SPANHCderived from [HC] together, a novel ship detection method SPANSDPis finally proposed to detect ships. Experiments performed on the airborne SAR (AIRSAR) L-Band and GF-3 C-Band data verify that 1) SPANSDPcan detect small ships more accurately than other state-of-the-art methods and 2) [HC] is more effective in improving ship detectors' detection performances in comparison with [CP].

The compact polarimetric (CP) imaging mode is a special, dual-polarization mode in which only one polarization is transmitted and two orthogonal polarizations are simultaneously used to measure the returns. The transmitting polarization can be an arbitrary elliptical wave, and therefore, theoretically, there are numerous possibilities of hybrid dual-pol modes. Since the scattering process is a function of the incident field and scattered field, when we use polarization to measure the backscattering process, radar measurement is polarization-dependent. This leads to difficulties in developing a unified interpretation for target scattering characterization in compact polarimetry, since the scattering models should be modified to take the fact of polarization dependency into account. In this letter, in the framework of a previously proposed CP formalism method, we develop a general model-based target decomposition technique that is applicable to the general CP mode. First, a general link between the fully polarimetric (FP) and the general CP scattering matrix elements is established. Second, the FP models corresponding to a single symmetric scattering mechanism and azimuthally symmetric volume scattering are mapped to the general CP case. Assuming that the 2-D backscatter is composed of a single scattering component and a volume scattering component, the decomposition is finally implemented by solving a quadratic equation. The C-band Shuttle Imaging Radar with Payload C/ X-synthetic aperture radar (SIR C/X-SAR) data are used in the experiment for demonstration.

Electromagnetic scattering and emission from 1-D rough surfaces with multiple elevations are studied using full-wave simulations. Both the root-mean-square (rms) heights and the surface length are large compared to the wavelength. A novel multilevel steepest decent-sparse matrix canonical grid (MLSD-SMCG) method is proposed to address limitations in the original SMCG. The uniform Nystrom method and neighborhood impedance boundary condition (NIBC) are also incorporated in solving the dual surface integral equations (SIEs) of the method of moments (MoM). Simulation results are illustrated at L-band for soil and ocean surfaces. The surface rms heights and lengths are up to 1.43 and 243.8 m corresponding to 6 and 1024 wavelengths at 1.26 GHz, respectively. For ocean surfaces, the wind speeds up to 20 m/s are considered, and the entire spectrum is included to capture all relevant surface length scales. Numerical results indicate the proposed approach is computationally efficient and accurate. Energy conservation checks in simulations are at $10^{-4}$ for ocean scattering and emission. Also, the effects of wind-driven roughness on ocean emissivity are further investigated using the proposed approach in terms of wind speed and observation angle for both polarizations.

We employ the second-order small slope approximation (SSA-II) and the method of moment (MoM) to investigate the roughness scale effect on ocean radar scattering in both three- (3-D) and two-dimensional (2-D) cases. Various roughness scales of the ocean surface are represented by truncating the Kudryavtsev wave spectrum. Criteria of full spectrum truncation are proposed for the numerical simulations of ocean scattering. Numerical results are illustrated in fully bistatic configuration at L- and C- bands. It is found that short waves with wavenumber larger than 316 rad/m have little effects on ocean scattering. The large-scale waves put more effects on scattering in the forward directions, particularly for large incidence angles. For numerical simulations of ocean scattering with incidence angle less than 60°, using small surface profiles with size about 1/6 of those accounting for full spectrum yields result with errors less than 2dB.

Accurate calculations of electromagnetic scattering and emission from ocean surfaces with large sizes and root-mean-square (RMS) heights are still a challenge for the fullwave numerical methods. The widely used sparse matrix canonical grid (SMCG) method is only efficient for small and moderate roughness surfaces. This is because of the limitation of approximations in the far-field expressions using Taylor series expansion. In this paper, a novel multilevel steepest decent-sparse matrix canonical grid (MLSD-SMCG) method is proposed to address this issue. The proposed approach is implemented with the method of moment (MoM) in solving the dual surface integral equations (SIEs). Simulation results are illustrated at L-band for ocean surfaces with wind speed up to 20 m/s. The entire spectra are involved to capture all scales of waves in simulations. Thus the surface root-mean-square (RMS) heights and lengths are up to 3.82 and 1024 wavelengths of 1.26 GHz, respectively. Numerical results indicate the proposed approach is computationally efficient and accurate. Energies in simulations are conserved to the order of 10-4for ocean scattering and emission at various wind speeds.

Classification is an important topic in synthetic aperture radar (SAR) image processing and interpretation. Markov Random Field (MRF) has been widely used for capturing the spatial-contextual information of the image. In this study, we first introduce two ways to construct the Wishart mixture model, and combine these models with MRF. Then, we use these models to implement pixel-based classification using a real PolSAR image. Finally, we analyze the experimental results and evaluate the robustness and applicability of the model mixture strategies.

Compact polarimetric interferometric synthetic aperture radar (CPolInSAR) has been widely used due to its low complexity and costs. But in the field of land cover classification, there are very few studies on CPolInSAR images. In this paper, a novel classification method is proposed to investigate the performance of CPolInSAR images for land cover classification. Specifically, the m - δ decomposition and the optimal polarimetric coherence coefficient are first employed to extract the features of CPolInSAR images, and then support vector machine (SVM) is utilized for classification. The experimental results show that 1) the optimal polarimetric coherence coefficient can be used to achieve higher accuracy especially in wood land and built-up areas; 2) CPolInSAR has a greater potential for land cover classification than PolSAR and compact PolSAR.

The scattering process is a function of the incident field and the scattered field. When we use polarization to measure the backscattered process, radar measurement is polarization dependent. This dependency causes difficulties in developing unified interpretations for scattering characterization in compact polarimetry because the compact polarimetric (CP) system could be with different transmit polarizations. In this study, in the framework of a previously proposed CP formalism method, we develop a general model-based target decomposition technique applicable to the general CP mode related to an arbitrary transmit wave polarization. Firstly, a general link between the fully polarimetric (FP) and CP scattering matrix elements is established. Secondly, the FP symmetric scattering model and the volume scattering model are mapped into the general CP case. By assuming that the 2-dimentional backscatter is composed of a single scattering mechanism and a volume scattering component, finally the decomposition can be implemented by solving a quadratic equation. We apply this method for oil spill detection and discrimination. C-band SIR-C/X-SAR data are used for validation to show the effectiveness of the general CP mode for ocean surface characterization.

In this paper, a new scattering difference parameter named as SDP is first constructed to characterize the relationship between polarimetric covariance matrix [C] and complete polarimetric covariance difference matrix [CP]. Then, by integrating ”1-SDP” and the power maximization synthesis detector (PMS) derived from [CP], a novel ship detection method OmSPcpis further developed. In order to demonstrate the performance of the proposed method, one AIRSAR L-Band Polarimetric SAR dataset with 22 ships is exploited. The experimental results show that, compared to other methods, OmSPcpcan hold a better ship detection accuracy.

The roughness scale effects on ocean radar scattering are studied using the second-order small slope approximation (SSA-II) and the method of moments (MoM). The KHCC03 spectrum is employed to represent 2-D and 1-D sea surfaces in the above two scattering methods, respectively. Criteria of full spectrum truncation are proposed for the numerical simulations of ocean scattering. Numerical results are illustrated in fully bistatic configuration at L- and C-bands. It is found that scattering at higher frequency is relatively more sensitive to the small-scale roughness but less sensitive to the large-scale roughness. At L- and C-bands, short waves with wavenumber larger than 316 rad/m have little effect on ocean scattering. The large-scale waves put more impacts on scattering in the forward directions, especially for large incidence angles. Other than the specular direction, the effects of large-scale roughness on ocean scattering are in general smaller at VV-pol than HH-pol. The bistatic scattering at cross polarizations is less sensitive to the roughness scale as compared to the copolarizations. For numerical simulations of ocean scattering with incidence angle less than 60°, using small surface profiles with size about 1/6 of those accounting for full spectrum yields results with errors less than 2 dB. Results also indicate that the incoherent parts dominate the scattered power from ocean surfaces with large-scale roughness.

Compared with optical images, polarimetric synthetic aperture radar (PolSAR) images usually maintain lower resolution. Ship targets in PolSAR images have fewer pixels than those in optical images. Therefore, architectures such as Faster R-CNN and its variations that focus more on target pixels may fail for PolSAR images. Due to limited data and the simple geometric structures of small ship targets, the R-CNN framework with a large neural network as its backbone easily overfits the training set. In this article, we propose a novel lightweight patch-to-pixel convolutional neural network (P2P-CNN) for ship detection via PolSAR images. P2P-CNN focuses on both the target and its surroundings. A patch of proper size that contains both a target and its surroundings is used as the input of the neural network to determine whether the pixel in the center of the patch belongs to the target. To utilize contextual semantic information at all scales, all feature maps from the top down are combined to improve the final result. Instead of conventional convolution, dilated convolutions are used in the proposed neural network to exponentially expand receptive fields without adding any model parameters. The proposed approach and comparative methods are tested and compared on PolSAR images in various environments under varying image resolution, target size, sea condition, sensor type, etc. The experimental results demonstrate that the proposed approach outperforms all the compared methods.

In this paper, we propose a novel ship detection method for polarimetric synthetic aperture radar (PolSAR) images via graph-based sparse manifold ranking. The main framework comprises a coarse-to-fine scheme. We employ the image intensity for prescreening and introduce graph-based sparse manifold ranking (GSMR) for discrimination. The distance between the ship and clutter in sparse code domain is explored. And the image elements in candidate region are ranked based on the prescreening priors and the new distance in graph labeling framework. The final detection result is produced with the ranking procedure. Experiments performed on two RADARSAT-2 images demonstrate the effectiveness and superiority of the proposed method.

In the past few years, land cover classification in polarimetric Synthetic Aperture Radar (SAR) images has been one of the research hotspots. The Wishart distribution has been developed to design the classical Wishart classifier. This classifier can classify each pixel by iteratively update the class centers. But the accuracy of the class center is affected by the misclassified points of the previous results, which has bad influence on the final result. Therefore, a local competitive strategy based on the Wishart classifier is proposed. In the proposed scheme, the class center is only updated in local area, which can reduce the impact of misclassified points. A real polarimetric image is used in experiments. Compared with other three methods, the proposed classification scheme achieves higher accuracy, and classification map is in better visual inspection.

Compact polarimetry (CP) is a dual-polarization hybrid imaging mode, which only provides partially polarized backscatter characteristics of scatterers. The compact polarimetric synthetic aperture radar (SAR) measures different fully polarimetric scattering elements when the illumination wave parameter varies. The backscattered wave is fully characterized by two channel elements, in which the channel ratio is important in identifying physical scattering mechanisms. In this study, we propose a new formalism method for general compact polarimetric data, such that the compact measurements are presented according to a same standard for describing target physical properties. Then based on this formalism, a channel ratio-based scattering characterization method is proposed, which is mathematically equal to the fully polarimetric ΔαB/αBmethod but without any scattering assumption. The formalism provides a natural way to characterize the target scattering behaviors. Experiments show that based on this formalism, the proposed method is convenient and efficient for compact target scattering interpretation.

Compact polarimetry is a dual-polarization hybrid imaging mode and provides partially backscatter characteristics of scatterers. The compact polarimetric (CP) synthetic aperture radar (SAR) measures the combinations of fully polarimetric (FP) scattering coefficients, which is dependent on the transmit-wave polarization state. The backscattered wave is fully described by a 2-D complex scattering vector, in which the dual-channel polarization ratio represents the vector nature for describing target polarimetric properties. However, due to the dependence of CP measurements on the transmit wave, the complex channel ratio is explained differently for the same scatterers under different observation modes, e.g., the channel ratio ρ = ± j under the left circular mode and ρ = ±1 under the linear π/4 mode for the canonical trihedral and dihedral scatterers. The explanation diversity is inconvenient for the use of CP data, resulting in nonunified target decomposition algorithms. In this study, first, we propose a new formalism method for the general CP SAR mode, such that all CP mode observables are described based on the same standard for target scattering characterization. This provides potential for developing unified CP decomposition algorithms. Then, based on this formalism, a polarization ratio-based scattering characterization method is proposed, which is mathematically equal to the FP ΛαB/αBmethod but without any scattering assumption. Theoretical analyses and discussions are provided. Experiments show that the proposed method is efficient for compact polarimetric scattering interpretation.

Superpixel segmentation has been commonly used in various image processing tasks such as object detection and image classification. In this paper, we propose a novel superpixel segmentation method based on a new distance function and superpixel seed updating strategy for polarimetric synthetic aperture radar (PolSAR) images. We initialize superpixel seeds with an expected number of superpixels. Then, we iteratively cluster the pixels based on the distance function and update the superpixel seeds based on the updating strategy. When the termination condition is reaches, we stop the iteration and obtain the superpixels. The experimental results based on RADARSAT-2 data demonstrate that our method is effective and achieves a better tradeoff between boundary adherence and compactness.

On 10 August 2016 China launched the GF-3, its first C-band polarimetric synthetic aperture radar (SAR) satellite, which was put into operation at the end of January, 2017. The GF-3 polarimetric SAR has many advantages such as high resolution and multi-polarization imaging capabilities. Polarimetric SAR can fully characterize the backscatter property of targets, and thus it is of great interest to explore the physical scattering mechanisms of terrain types, which is very important in interpreting polarimetric SAR imagery and for its further usages in Earth observation. Both Radarsat-2 and Gaofen-3 satellites operate in C-band and are designed for similar missions. In this study, we compare the performances of fully polarimetric Gaofen-3 and Radarsat-2 data in representation of terrain types. By using several classic polarimetric features and the iterative Wishart-distribution based classifier, Gaofen-3 and Radarsat-2 data are compared in the aspects of polarimetric property preservation and capability of terrain classification.

The optimal receiving technique of target backscattered signal in radar polarimetry has been successfully applied in many applications. In this study, we further develop the change detector which was proposed based on the optimization of polarimetric contrast of radar received powers. The change detector used for multi-temporal polarimetric synthetic aperture radar (SAR) analysis was proposed with the purpose to seek a pair of optimal polarization states such that fluctuation between the observables collected at different times can be minimized. In this study, the change detector is simplified. Simulated polarimetric data is used to verify the simplification. Real polarimetric SAR data sets are used for demonstration and the performance is compared with the likelihood ratio test (LRT) method implemented in PolSARpro.

Change detection is an important issue in many applications. In this study, we propose a new change detector for multi-temporal polarimetric synthetic aperture radar (SAR) images. The new detector is based on the optimization of polarimetric contrast minimization. The optimized solution of the contrast model is also given. Experiments are performed on two RADARSAT-2 data sets and results show agreement with land changes.

This letter presents a multiscale edge detection method for multilook polarimetric synthetic aperture radar (PolSAR) images based on the nonsubsampled contourlet transform (NSCT). The NSCT can provide flexible multiscale and directional decomposition. In the multiscale decomposition, the coefficients of the nonsubsampled pyramid in the NSCT are calculated via maximizing the polarimetric contrast between the adjacent subband levels, instead of using the difference of the adjacent subbands as used in the additive noise model. By this way, we make the NSCT applicable to PolSAR data and multiband data. Then, the edges are detected in the NSCT domain based on a fusion of the directional subband coefficients at different scales. Experimental results with both simulated and real PolSAR data show that the present approach is robust to noise and the extracted edges are complete and continuous.

In this letter, a constant false-alarm rate line detector for polarimetric synthetic aperture radar (Pol-SAR) images is presented based on Wilks' test statistic, which can be used to test the equality of two covariance matrices following the complex Wishart distribution. Due to the two-tailed nature, Wilks' test statistic can detect both bright and dark features. The probability distribution of the proposed detector is derived and verified using simulated Pol-SAR images. Experimental results on both simulated and measured Pol-SAR data show the effectiveness of the proposed detector and its superiority over the traditional detector based on the Wishart likelihood-ratio test statistic.

This letter proposes a harbor detection method by extracting specific paralleled contours of harbors in polarimetric synthetic aperture radar (SAR) images. First, land and sea are partitioned by the region-based level set segmentation algorithm which is applied to the extraction of the volume scattering component. Then, parallel curves are modeled by a series of parallel line segments. They are detected by finding out and connecting paralleled line segments given certain criteria. The detected parallel curves are merged according to a distance measurement so that the harbor contour is finally detected. RADARSAT-2 polarimetric SAR data covering an area in Singapore are used to test the harbor detection method. Experimental results show that the method improves sea-land segmentation accuracy and extracts all parallel curves along the coast correctly. The method enables an accurate detection of harbors.

One basic issue of importance in polarimetric synthetic aperture radar (SAR) imagery is the identification and separation of target scattering mechanisms. Physical scattering behaviors can be characterized by polarimetric parameters from the second-order statistical observables. The average copolarization phase difference, amplitude ratio, and target coherence are important fundamental parameters for identifying scattering mechanisms. However, the individual usages of these parameters could not describe both the scattering mechanisms and the depolarization. In this paper, a new approach is proposed for scattering characterization by exploring the information contained in these three parameters. First, by assuming reflection symmetry, a new parameter is proposed for the first time to measure the scattering randomness. Then, in combination with the scattering ratio (defined by the ratio of T22+ T33to T11), a classification plane is proposed to classify target scattering mechanisms. A validation test for this new approach is performed with three RADARSAT-2 polarimetric data sets acquired over two study areas: the San Francisco Bay area and Fuzhou, China. Results show that the new approach is very promising for distinguishing orientated targets (with respect to the radar azimuth direction) in urban areas from natural scatterers such as forests, and it also shows that the new method is robust for analyzing multitemporal polarimetric SAR data.

Scattering mechanism identification is an important issue in polarimetric SAR imagery. In this study, we propose a new method for target scattering mechanism classification for the first time. Based on several typical scattering models, including the X-Bragg scattering model, Fresnel reflection model, and volume scattering models, a new polarimetric scattering classification diagram is proposed. It consists of two parameters, i.e., the Bragg alpha angle and a new parameter which is proposed for the first time and can be used to represent backscattering randomness. Results are demonstrated and validated on RADARSAT-2 data of the San Francisco area to illustrate its effectiveness.

This study focuses on the development and evaluation of the generalized optimization of polarimetric contrast enhancement (GOPCE) model to discriminate between forested and nonforested areas. The main objective is to investigate the performance of the GOPCE method for forest mapping and to assess the potential of different polarimetric parameters for forest representation. We make two modifications to the original GOPCE method. First, by comparing behaviors of different polarimetric parameters, the GOPCE model is modified. Then, linear discriminant analysis is employed for further optimization of the target contrast. Forest/nonforest discrimination results are demonstrated on L-band fully polarimetric ALOS-1/PALSAR data acquired over a pilot study area in northeastern Tasmania, Australia, where the main forest type is eucalypt forests. Two other forest classification approaches (i.e., support vector machine and canonical variate analysis) are also tested for comparison. The final results obtained from the modified GOPCE model with the generalized Fisher criterion can improve the forest/nonforest discrimination accuracy.

Ocean surveillance is an important application in synthetic aperture radar (SAR) imagery. Polarimetric SAR (Pol-SAR) provides the multi-channel scattering information and hence is very promising for earth observation. Recently, the dual polarimetric SAR especially the hybrid dual-pol mode (i.e., compact polarimetry) has been given a growing concern. It can provide wide imaging swath coverage with reduced system complexity and thus is of great potential for maritime surveillance applications. Compact decomposition methods have been widely evaluated and applied to the land cover applications. In this study, we investigate the performances of compact decompositions for ship detection. The surveillance scenario is focused on the coastal areas, where low wind regions (LWR) and small ships with low backscattered intensity often appear in Pol-SAR imagery. The compact m-δ and m-χ decompositions are included to show the importance of the contribution of total backscattered energy in ship detection. Results are demonstrated on two RADARSAT-2 acquisitions over Tianjin port and Dalian port, China.

In this letter, a new method for Kennaugh matrix decomposition is proposed, and a new norm for the Kennaugh matrix is defined. The Kennaugh matrix is decomposed into two parts: The first is a coherent target matrix, and the second is a residual matrix with minimum norm. The properties of the extracted coherent target are discussed, and an application of the extracted coherent target is implemented. In this application, an incoherent image is converted into a coherent image. Single-look sphere-diplane-helix decomposition is then performed. An experiment on Airborne SAR polarimetric data over San Francisco has been carried out, thus demonstrating the effectiveness of the application.

This paper investigates the application of a level set method for the automated multiphase segmentation of multiband and polarimetric synthetic aperture radar (SAR) images. The level set formulation is used to form an energy functional that includes the image statistical information defined on active contours. In addition to the classical Wishart/Gaussian distribution for locating region boundaries, edge information is incorporated into the energy functional to improve the performance of polarimetric data segmentation. An active contour model with an edge indicator is proposed by assuming that the image boundary term follows a Gibbs prior. An empirical parameter setting criterion is developed to ensure that the components of the energy functional are in proper proportion. We then investigate the multiphase extension for energy minimization, and we use a piecewise constant model to embed the proposed active contour model. Synthetic and real multiband polarimetric SAR data are used for verification. The experiments show that our method is superior to another level set method based on the Wishart/Gaussian distribution, in which SAR edge information is not included, particularly for discriminating among low-contrast regions. Furthermore, results also show that segmentation is improved when multiband data are used in the level set framework.

In this paper, a new parameter called Bragg energy proportion is introduced for spill oil detection, based on Freeman decomposition model. Bragg energy proportion is small in spill region and it is large in sea clutter region or Oleyl Alcohol (OLA) region. So, this parameter can be used to detect oil spill from sea clutter and meanwhile it reduces the OLA false-alarm effectively. In addition, the parameter is derived by HH\VV dual-polarized SAR data. Using C band SIR-C SAR data, the authors demonstrate the effectiveness of the proposed method.

Polarization ratio and co-polarized phase information are very important for polarimetric synthetic aperture radar (SAR) image interpretation, especially in the area where a single scattering mechanism is dominant. In this study, a new method including both the parameters for scattering characterization is proposed based on the extended Bragg scattering (X-Bragg) model. The theoretical analysis, which is based on the first-order Bragg scattering and the dihedral scattering models, is consistent with the demonstration of real polarimetric SAR data. The proposed method is evaluated on the entropy/alpha plane, showing its promise for distinguishing between different target scatterings. In experiment, L-band ALOS/PALSAR fully polarimetric data over Tasmania, Australia are used for illustrating the effectiveness of this method for forest/nonforest discrimination. Results from Cloude-Pottier's decomposition theorem are also given for comparison.