Original Synthetic Aperture Radar Image Research Articles

Improving SAR Target Recognition Performance Using Multiple Preprocessing Techniques.

The synthetic aperture radar (SAR) image preprocessing techniques and their impact on target recognition performance are researched. The performance of SAR target recognition is improved by composing a variety of preprocessing techniques. The preprocessing techniques achieve the effects of suppressing background redundancy and enhancing target characteristics by processing the size and gray distribution of the original SAR image, thereby improving the subsequent target recognition performance. In this study, image cropping, target segmentation, and image enhancement algorithms are used to preprocess the original SAR image, and the target recognition performance is effectively improved by combining the above three preprocessing techniques. On the basis of image enhancement, the monogenic signal is used for feature extraction and then the sparse representation-based classification (SRC) is used to complete the decision. The experiments are conveyed on the moving and stationary target acquisition and recognition (MSTAR) dataset, and the results prove that the combination of multiple preprocessing techniques can effectively improve the SAR target recognition performance.

Multiset Canonical Correlations Analysis of Bidimensional Intrinsic Mode Functions for Automatic Target Recognition of SAR Images.

A novel feature generation algorithm for the synthetic aperture radar image is designed in this study for automatic target recognition. As an adaptive 2D signal processing technique, bidimensional empirical mode decomposition is employed to generate multiscale bidimensional intrinsic mode functions from the original synthetic aperture radar images, which could better capture the broad spectral information and details of the target. And, the combination of the original image and decomposed bidimensional intrinsic mode functions could promisingly provide more discriminative information for correct target recognition. To reduce the high dimension of the original image as well as bidimensional intrinsic mode functions, multiset canonical correlations analysis is adopted to fuse them as a unified feature vector. The resultant feature vector highly reduces the high dimension while containing the inner correlations between the original image and decomposed bidimensional intrinsic mode functions, which could help improve the classification accuracy and efficiency. In the classification stage, the support vector machine is taken as the basic classifier to determine the target label of the test sample. In the experiments, the 10-class targets in the moving and stationary target acquisition and recognition dataset are classified to investigate the performance of the proposed method. Several operating conditions and reference methods are setup for comprehensive evaluation.

Target Recognition of SAR Images Based on SVM and KSRC.

A synthetic aperture radar (SAR) target recognition method combining linear and nonlinear feature extraction and classifiers is proposed. The principal component analysis (PCA) and kernel PCA (KPCA) are used to extract feature vectors of the original SAR image, respectively, which are classical and reliable feature extraction algorithms. In addition, KPCA can effectively make up for the weak linear description ability of PCA. Afterwards, support vector machine (SVM) and kernel sparse representation-based classification (KSRC) are used to classify the KPCA and PCA feature vectors, respectively. Similar to the idea of feature extraction, KSRC mainly introduces kernel functions to improve the processing and classification capabilities of nonlinear data. Through the combination of linear and nonlinear features and classifiers, the internal data structure of SAR images and the correspondence between test and training samples can be better investigated. In the experiment, the performance of the proposed method is tested based on the MSTAR dataset. The results show the effectiveness and robustness of the proposed method.

Adversarial Examples for CNN-Based SAR Image Classification: An Experience Study

Synthetic aperture radar (SAR) has all-day and all-weather characteristics and plays an extremely important role in the military field. The breakthroughs in deep learning methods represented by convolutional neural network (CNN) models have greatly improved the SAR image recognition accuracy. Classification models based on CNNs can perform high-precision classification, but there are security problems against adversarial examples (AEs). However, the research on AEs is mostly limited to natural images, and remote sensing images (SAR, multispectral, etc.) have not been extensively studied. To explore the basic characteristics of AEs of SAR images (ASIs), we use two classic white-box attack methods to generate ASIs from two SAR image classification datasets and then evaluate the vulnerability of six commonly used CNNs. The results show that ASIs are quite effective in fooling CNNs trained on SAR images, as indicated by the obtained high attack success rate. Due to the structural differences among CNNs, different CNNs present different vulnerabilities in the face of ASIs. We found that ASIs generated by nontarget attack algorithms feature attack selectivity, which is related to the feature space distribution of the original SAR images and the decision boundary of the classification model. We propose the sample-boundary-based AE selectivity distance to successfully explain the attack selectivity of ASIs. We also analyze the effects of image parameters, such as image size and number of channels, on the attack success rate of ASIs through parameter sensitivity. The experimental results of this study provide data support and an effective reference for attacks on and the defense capabilities of various CNNs with regard to AEs in SAR image classification models.

SAR image despeckling method based on improved Frost filtering

In this paper, a novel despeckling algorithm for synthetic aperture radar (SAR) images based on improved Frost filtering is proposed. A reconstructed decay factor which can better characterize the homogeneous and edge regions of SAR image is added into the Frost filter model by using the Lee filter coefficient to implement adaptively controlling the degree of filtering in different regions. The three-stage filtering strategy is performed for effective despeckling and edge preservation. The SAR image is pre-filtered by using the Frost filtering of the small window to obtain local statistics, and then the original SAR image is refine filtered twice utilizing these local statistical parameters. The two values are, respectively, biased toward smoothing and edge preservation by adjusting the variable constant and the window scales. Finally, the final filtering result is obtained by weighting the two filtering results with the reconstructed decay factor we proposed. The experiments demonstrate that compared with advanced approaches, the proposed method improves both despeckling and edge preservation ability.

Target classification of SAR images using nonlinear correlation information entropy

A synthetic aperture radar (SAR) target classification method is proposed by selecting deep features learned by convolutional neural network (CNN) using the nonlinear correlation information entropy (NCIE). The deep feature maps are learned from original SAR images by CNN. Considering the possible indiscrimination and redundancy, all the deep feature maps from different layers are analyzed by NCIE to select out those discriminative ones. The selected feature maps share high inner correlations, which are jointly classified based on joint sparse representation (JSR). Finally, the target label is determined based on the reconstruction errors from JSR. The famous moving and stationary target acquisition and recognition dataset are employed to setup the standard operating condition and several extended operating conditions, including configuration variants, depression angle differences, noise corruption, and partial occlusion, to comprehensively evaluate the performance of the proposed method. The results confirm the superior effectiveness and robustness of the proposed method in comparison with some state-of-the-art SAR target classification methods.

SAR Speckle Removal Using Hybrid Frequency Modulations

Synthetic aperture radar (SAR) images often interfere with speckle artifacts that have a great impact on subsequent processing and analysis operations. To remove speckle artifacts, this article introduces a hybrid denoising approach by using a convolutional neural network (CNN) and consistent cycle spinning (CCS) in the nonsubsample shearlet transform (NSST) domain. First, we apply NSST to a noisy SAR image to gain low- and high-frequency coefficients. Second, we adopt a learned deep CNN model to eliminate the speckle noise in the low-frequency coefficients, which retains more contour information. Third, we employ CCS to enhance the high-frequency coefficients, which preserves more details of the original SAR image. Finally, we obtain the denoised image by using inverse NSST applied to the denoised coefficients. Compared with state-of-the-art algorithms, the results of the experiment indicate that our method not only achieves better speckle removal performance but also maintains more detailed information retention.

High-resolution SAR images segmentation using NSCT denoising and QIGA based parameters selection of PCNN model

The high-resolution synthetic aperture radar (SAR) images usually contain inhomogeneous coherent speckle noises. For the high-resolution SAR image segmentation with such noises, the conventional methods based on pulse coupled neural networks (PCNN) have to face heavy parameters with a low efficiency. In order to solve the problems, this paper proposes a novel SAR image segmentation algorithm based on non-subsampling Contourlet transform (NSCT) denoising and quantum immune genetic algorithm (QIGA) improved PCNN models. The proposed method first denoising the SAR images for a pre-processing based on NSCT. Then, by using the QIGA to select parameters for the PCNN models, such models self-adaptively select the suitable parameters for segmentation of SAR images with different scenes. This method decreases the number of parameters in the PCNN models and improves the efficiency of PCNN models. At last, by using the optimal threshold to binary the segmented SAR images, the small objects and large scales from the original SAR images will be segmented. To validate the feasibility and effectiveness of the proposed algorithm, four different comparable experiments are applied to validate the proposed algorithm. Experimental results have shown that NSCT pre-processing has a better performance for coherent speckle noises suppression, and QIGA-PCNN model based on denoised SAR images has an obvious segmentation performance improvement on region consistency and region contrast than state-of-the-arts methods. Besides, the segmentation efficiency is also improved than conventional PCNN model, and the level of time complexity meets the state-of-the-arts methods. Our proposed NSCT+QIGA-PCNN model can be used for small object segmentation and large scale segmentation in high-resolution SAR images. The segmented results will be further used for object classification and recognition, regions of interest extraction, and moving object detection and tracking.

Representation of BVMD features via multitask compressive sensing for SAR target classification

ABSTRACT This letter develops a synthetic aperture radar (SAR) target classification method based on bidimensional variational mode decomposition (BVMD) and multitask compressive sensing (MTCS). BVMD is employed to decompose SAR images to exploit the time-frequency properties of the described targets. The MTCS is used to jointly classify the original SAR image and its BVMD components. So, the merits of BVMD and MTCS can be combined in the proposed method. Finally, based on the reconstruction errors, the target label can be decided. The Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset is used to set up experimental conditions to test the proposed method. By comparison with several reference methods from published works, the effectiveness and robustness of the proposed method can be confirmed.

Target Classification From SAR Imagery Based on the Pixel Grayscale Decline by Graph Convolutional Neural Network

The target classification from synthetic aperture radar (SAR) imagery is entering a bottleneck stage for the extensive use of a deep learning technology. Researchers have deployed various deep neural networks to extract the target features from the original SAR image in Euclidean space, which requires a large number of training data and cost lots of time to train the deep neural networks well generalized. Aiming at this problem, this letter introduces a novel method of target classification from SAR imagery based on the target pixel grayscale decline by a graph representation, which is different from the conventional deep learning methods so far. We separate the whole grayscale interval of one SAR image into several subintervals and assign a node to represent each pixel with the declined order of pixel grayscale in the subinterval. Then, a graph structure could be constructed to transform the raw SAR image from Euclidean data to graph-structured data. Finally, we construct a graph convolutional neural network to extract the features of graph-structured data we constructed previously and output the target classification result. The experiment result on the MSTAR dataset shows that our method achieved the average classification accuracy with 100%, which surpasses all the state-of-the-art methods for the first time in SAR automatic target recognition field.

Saliency-Guided Single Shot Multibox Detector for Target Detection in SAR Images

The single shot multibox detector (SSD), a proposal-free method based on convolutional neural network (CNN), has recently been proposed for target detection and has found applications in synthetic aperture radar (SAR) images. Moreover, the saliency information reflected in the saliency map can highlight the target of interest while suppressing clutter, which is beneficial for better scene understanding. Therefore, in this article, we propose a saliency-guided SSD (S-SSD) for target detection in SAR images, in which we effectively integrate the saliency into the SSD network not only to suggest where to focus on but also to improve the representation capability in complex scenes. The proposed S-SSD contains two separated convolutional backbone subnetwork architectures, one with the original SAR image as input to extract features, and the other with the corresponding saliency map obtained from the modified Itti’s method as input to acquire refined saliency information under supervision. In addition, the dense connection structure, instead of the plain structure used in original SSD, is applied in the two convolutional backbone architectures to utilize multiscale information with fewer parameters. Then, for integrating saliency information to guide the network to emphasize informative regions, multilevel fusion modules are utilized to merge the two streams into a unified framework, thereby making the whole network end-to-end jointly trained. Finally, the convolutional predictors are used to predict targets. The experimental results on the miniSAR real data demonstrate that the proposed S-SSD can achieve better detection performance than state-of-the-art methods.

Potential of texture from SAR tomographic images for forest aboveground biomass estimation

Synthetic Aperture Radar (SAR) texture has been demonstrated to have the potential to improve forest biomass estimation using backscatter. However, forests are 3D objects with a vertical structure. The strong penetration of SAR signals means that each pixel contains the contributions of all the scatterers inside the forest canopy, especially for the P-band. Consequently, the traditional texture derived from SAR images is affected by forest vertical heterogeneity, although the influence on texture-based biomass estimation has not yet been explicitly explored. To separate and explore the influence of forest vertical heterogeneity, we introduced the SAR tomography technique into the traditional texture analysis, aiming to explore whether TomoSAR could improve the performance of texture-based aboveground biomass (AGB) estimation and whether texture plus tomographic backscatter could further improve the TomoSAR-based AGB estimation. Based on the P-band TomoSAR dataset from TropiSAR 2009 at two different sites, the results show that ground backscatter variance dominated the texture features of the original SAR image and reduced the biomass estimation accuracy. The texture from upper vegetation layers presented a stronger correlation with forest biomass. Texture successfully improved tomographic backscatter-based biomass estimation, and the texture from upper vegetation layers made AGB models much more transferable between different sites. In addition, the correlation between texture indices varied greatly among different tomographic heights. The texture from the 10 to 30 m layers was able to provide more independent information than the other layers and the original images, which helped to improve the backscatter-based AGB estimation.

A New Framework for Automatic Airports Extraction from SAR Images Using Multi-Level Dual Attention Mechanism

The detection of airports from Synthetic Aperture Radar (SAR) images is of great significance in various research fields. However, it is challenging to distinguish the airport from surrounding objects in SAR images. In this paper, a new framework, multi-level and densely dual attention (MDDA) network is proposed to extract airport runway areas (runways, taxiways, and parking lots) in SAR images to achieve automatic airport detection. The framework consists of three parts: down-sampling of original SAR images, MDDA network for feature extraction and classification, and up-sampling of airports extraction results. First, down-sampling is employed to obtain a medium-resolution SAR image from the high-resolution SAR images to ensure the samples (500 × 500) can contain adequate information about airports. The dataset is then input to the MDDA network, which contains an encoder and a decoder. The encoder uses ResNet_101 to extract four-level features with different resolutions, and the decoder performs fusion and further feature extraction on these features. The decoder integrates the chained residual pooling network (CRP_Net) and the dual attention fusion and extraction (DAFE) module. The CRP_Net module mainly uses chained residual pooling and multi-feature fusion to extract advanced semantic features. In the DAFE module, position attention module (PAM) and channel attention mechanism (CAM) are combined with weighted filtering. The entire decoding network is constructed in a densely connected manner to enhance the gradient transmission among features and take full advantage of them. Finally, the airport results extracted by the decoding network were up-sampled by bilinear interpolation to accomplish airport extraction from high-resolution SAR images. To verify the proposed framework, experiments were performed using Gaofen-3 SAR images with 1 m resolution, and three different airports were selected for accuracy evaluation. The results showed that the mean pixels accuracy (MPA) and mean intersection over union (MIoU) of the MDDA network was 0.98 and 0.97, respectively, which is much higher than RefineNet and DeepLabV3. Therefore, MDDA can achieve automatic airport extraction from high-resolution SAR images with satisfying accuracy.

Selection of Multi-Level Deep Features via Spearman Rank Correlation for Synthetic Aperture Radar Target Recognition Using Decision Fusion

Convolutional neural networks (CNN) now become one of the most popular methods in synthetic aperture radar (SAR) target recognition. To fully exploit the deep features learned by CNN, this paper considers all the feature maps from different convolution layers. At each layer, the Spearman rank correlation is employed to evaluate the similarities between the feature maps and original SAR image. A certain proportion of feature maps with high similarities are selected and jointly represented based on the joint sparse representation (JSR) model. For the reconstruction error vectors from different layers, they are combined based on linear weighting using a random weight matrix. The fused reconstruction errors are analyzed to form a decision value for target recognition. The feature selection chooses the robust features and JSR considers the inner correlations between the feature maps from the same layer. In addition, the linear weighting using the random weight matrix could statistically reveal the correlations between the test sample and a certain training class. Therefore, the overall effectiveness and robustness of the proposed method can be enhanced. By performing experiments on the moving and stationary target acquisition and recognition (MSTAR) dataset, the proposed method could achieve a very high average recognition rate of 99.32% for ten classes of ground targets under the standard operating condition (SOC). Furthermore, under the extended operating conditions (EOCs) like configuration differences, depression angle differences, noise corruption, and partial occlusion, the proposed could also achieve superior robustness over some state-of-the-art SAR target recognition methods.

A Novel Salient Feature Fusion Method for Ship Detection in Synthetic Aperture Radar Images

Ship detection of synthetic aperture radar (SAR) images is one of the research hotspots in the field of marine surveillance. Fusing salient features to detection network can effectively improve the precision of ship detection. However, how to effectively fuse the salient features of SAR images is still a difficult task. In this paper, to improve the ship detection precision, we design a novel one-stage ship detection network to fuse salient features and deep convolutional neural network (CNN) features. Firstly, a saliency map extraction algorithm is proposed. The algorithm is applied to generate saliency map by using multi-scale pyramid features and frequency domain features. Secondly, the backbone of the ship detection network contains a two-stream network. The upper-stream network uses the original SAR image as input to extract multi-scale deep CNN features. The lower-stream network uses the corresponding saliency map as input to acquire multi-scale salient features. Thirdly, for integrating the salient features to deep CNN features, a novel salient feature fusion method is designed. Finally, an improved bi-directional feature pyramid network is applied to the ship detection network for reducing the computational complexity and network parameters. The proposed methods are evaluated on the public ship detection dataset and the experimental results shows that it can make a significant improvement in the precision of SAR image ship detection.

A Novel Attention Fully Convolutional Network Method for Synthetic Aperture Radar Image Segmentation

As an important step of synthetic aperture radar image interpretation, synthetic aperture radar image segmentation aims at segmenting an image into different regions in terms of homogeneity. Because of the deficiency of the labeled samples and the existence of speckling noise, synthetic aperture radar image segmentation is a challenging task. We present a new method for synthetic aperture radar image segmentation in this article. Due to the large size of the original synthetic aperture radar image, we first divide the input image into small slices. Then the image slices are input to the attention-based fully convolutional network for obtaining the segmentation results. Finally, the fully connected conditional random field is adopted for improving the segmentation performance of the network. The innovations of our method are as follows: 1) The attention-based fully convolutional network is embedded with the multiscale attention network which is capable of enhancing the extraction of the image features through three strategies, namely, multiscale feature extraction, channel attention extraction, and spatial attention extraction. 2) We design a new loss function for the attention fully convolutional network by combining Lovasz-Softmax and cross-entropy losses. The new loss allows us to simultaneously optimize the intersection over union and the pixel classification accuracy of the segmentation results. The experiments are performed on two airborne synthetic aperture radar image databases. It has been proved that our method is superior to other state-of- the-art image segmentation approaches.

Target Recognition in SAR Image by Joint Classification of Target Region and Shadow

Background: A synthetic aperture radar (SAR) automatic target recognition (ATR) method is proposed in this paper via the joint classification of the target region and shadow. Methods: The elliptical Fourier descriptors (EFDs) are used to describe the target region and shadow extracted from the original SAR image. In addition, the relative positions between the target region and shadow are represented by a constructed feature vector. The three feature vectors complement each other to provide more comprehensive descriptions of the target’s physical properties, e.g., sizes and shape. In the classification stage, the three feature vectors are jointly classified based on the joint sparse representation (JSR). JSR is a multi-task learning algorithm, which can not only represent each component properly but also exploit the inner correlations of different components. Finally, the target type is determined to the class with the minimum reconstruction error. Results: Experiments have been conducted on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset. The proposed method achieves a high recognition accuracy of 96.86% for 10-class recognition problem under the standard operating condition (SOC). Moreover, robustness of the proposed method is also superior over the reference methods under the extended operating conditions (EOCs) like configuration variance, depression angle variance, and noise corruption. Conclusion: Therefore, the effectiveness and robustness of the proposed method can be quantitatively demonstrated by the experimental results.

Recognition of targets in SAR images using joint classification of deep features fused by multi-canonical correlation analysis

ABSTRACTThis letter proposes a synthetic aperture radar (SAR) target recognition method via joint classification of deep features fused by multi-canonical correlation analysis (MCCA). A convolutional neural network (CNN) is designed for feature learning from original SAR images. For the multiple feature maps from different convolution layers, they are fused based on the MCCA to maintain the relevance while eliminating the redundancy. Afterwards, the joint sparse representation (JSR) is employed to jointly represent the fused deep feature vectors from different convolution layers under the constraint of their inner correlations. Based on the reconstruction errors from JSR, the target label can be classified. The proposed method can make full use of the multi-level deep features by using the correlations among the same layer and between different layers. Experiments are investigated on the Moving and Stationary Target Acquisition and Recognition (MSTAR) data set and the results confirm the performance of the proposed method.

SAR image change detection based on deep denoising and CNN

The intrinsic noise of synthetic aperture radar (SAR) images has a big influence to the image processing performance, especially in change detection (CD). Image denoising is an important branch of image restoration which aims at enhancing the quality of images. The detection accuracy of CD depends greatly on the quality of red difference image (DI), therefore image denoising can be regarded as a vital step in SAR CD. However, few researches focused on this problem. In this study, an end-to-end deep denoising model is first designed to remove the noise of SAR images. With the help of abundant simulated SAR images, deep denoising model is trained effectively to estimate the noise component. Then clean image can be achieved by removing this noise component from the original SAR image. After denoising, the new image pair will generate a clean DI. At last, DI is classified into changed and unchanged areas by a three-layer Convolutional Neural Network (CNN). Three real SAR image pairs demonstrate the effectiveness of the proposed method.

An Energy-Based SAR Image Segmentation Method with Weighted Feature

To extract more structural features, which can contribute to segment a synthetic aperture radar (SAR) image accurately, and explore their roles in the segmentation procedure, this paper presents an energy-based SAR image segmentation method with weighted features. To precisely segment a SAR image, multiple structural features are incorporated into a block- and energy-based segmentation model in weighted way. In this paper, the multiple features of a pixel, involving spectral feature obtained from original SAR image, texture and boundary features extracted by a curvelet transform, form a feature vector. All the pixels’ feature vectors form a feature set of a SAR image. To automatically determine the roles of the multiple features in the segmentation procedure, weight variables are assigned to them. All the weight variables form a weight set. Then the image domain is partitioned into a set of blocks by regular tessellation. Afterwards, an energy function and a non-constrained Gibbs probability distribution are used to combine the feature and weight sets to build a block-based energy segmentation model with feature weighted on the partitioned image domain. Further, a reversible jump Markov Chain Monte Carlo (RJMCMC) algorithm is designed to simulate from the segmentation model. In the RJMCMC algorithm, three move types were designed according to the segmentation model. Finally, the proposed method was tested on the SAR images, and the quantitative and qualitative results demonstrated its effectiveness.