Linbin Zhang

Synthetic Aperture Radar (SAR) is an active remote sensing system that utilizes radar to produce images of the Earth's surface. Due to its ability to operate under diverse weather conditions and throughout the day, SAR has gained significant attention in both civilian and military domains. The utilization of multi-view SAR sequences enables the acquisition of a more comprehensive range of information than single image, and facilitates adaptation to diverse scenarios, thereby enhancing the ability to accommodate variations in samples. Drawing inspiration from the Transformer architecture, this paper proposes a multi-view SAR target recognition method, called Res-Xformer, that not only deconstructs the deep learning procedure into single image feature extraction and sequence feature fusion, but also divides the task of sequence feature extraction into sequence information fusion and feature channel fusion. Different from the Transformers focusing on the attention mechanism to fuse sequence information, alternative fusion methods such as multi-layer perceptron (MLP) and pooling are also proposed in this study. Experimental results using Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset demonstrate that the proposed method performs well across various operational conditions, with MLP and pooling as sequence token mixers yielding comparable performance to attention mechanism.

The data distribution of synthetic aperture radar (SAR) vehicle targets in the actual missions is often imbalanced. However, the recent algorithms for SAR target recognition are designed either under abundant samples, or the situation of few labeled samples among all the categories. These cases all avoid facing the difficulties of imbalanced data distribution, i.e . the difference between the number of labeled samples among categories is huge. The samples in the majority classes will get more chances to be learnt by the deep neural network, which impedes the regular algorithms from achieving a high recognition rate. In this letter, a design guideline for imbalance loss and an example of data distribution (DD) loss based on the guideline is proposed, which provides an extremely effective way of handling the problem of imbalanced SAR target recognition. The DD loss takes the sample distribution and the data quantity of SAR vehicle targets into consideration. It can cause images with fewer samples in their categories to decrease more gradients proportionally. Moreover, the proposed DD loss adds no more burden to the networks and compared to other imbalanced algorithms with complex processes, the DD loss can be conducted easily. Plenty of experiments, which involve two various kinds of imbalanced datasets, are implemented and the proposed DD loss shows excellent performance among these imbalanced datasets. When there are only 40 labeled samples in minority categories, the DD loss can achieve over 95% in nine different cases, which exceeds other methods and losses of at least 7%.

Synthetic aperture radar (SAR) automatic target recognition (ATR) has attracted extensive attention and achieved satisfactory results. However, most SAR ATR methods follow the closed-set assumption, which assumes that all target classes in the test set have been contained by the training set. In actual scenarios, it may encounter the target classes that are not included in the training set, and it presents a challenge for SAR ATR. To tackle this issue, this letter proposes an open-set recognition method based on joint training of class-specific sub-dictionary learning. First, joint training is used to optimize the sub-dictionary learning process, and it could significantly enhance the discriminative ability of these sub-dictionaries. Second, the reconstruction errors of the targets on each sub-dictionary are calculated. These errors can be split into matched and no-matched errors. Third, extreme value theory (EVT) is employed to model the matched and no-matched errors of each class, which could determine the class boundaries. Finally, for the target to be recognized, its reconstruction errors on each sub-dictionary are calculated individually. The class of this target can be determined by comparing the errors with these class boundaries. Our method achieved an accuracy of 87.22–94.02 and an F1 score of 88.03–90.25 in multiple experiments on moving and stationary target automatic recognition (MSTAR) dataset. Compared with several state-of-the-art methods, it has better accuracy and robustness.

SAR target classification is an important application in SAR image interpretation. In practical applications, the battlefield is open and dynamic, and the SAR target classification model often encounters the targets of unknown classes. However, most of the existing SAR target classification methods follow the close-set assumption. It makes them only classify several fixed classes of targets and can’t deal with the targets from unknown classes. To this end, this paper proposes a novel SAR target classification method. This method can not only classify the targets from known classes and search targets from unknown classes but also incrementally update the classification model with these unknown class targets. Specifically, an autoencoder improved by MS-SSIM (multi-scale structural similarity) loss is utilized to extract targets’ features, and it can better utilize the structural information in SAR images. Next, the classifier based on EVT (Extreme Value Theorem) is established, which can classify the known class targets and search the unknown class targets. Then, we perform improved model reduction on the established classifier. This operation could speed up the model and prepare for incremental learning. Finally, after manually labeling those unknown class targets, the classifier is updated with these data in incremental form. Experimental results on the MSTAR (Moving and Stationary Target Automatic Recognition) dataset indicate that, compared with the state-of-the-art methods, our proposed method has better performance in open set recognition and incremental learning.

Machine learning methods for synthetic aperture radar (SAR) image automatic target recognition (ATR) can be divided into two main types: traditional methods and deep learning methods. The deep learning methods can learn the high-dimensional features of the target directly, and usually obtain high target recognition accuracy. However, they lack full consideration of SAR targets’ inherent characteristics resulting in poor generalization and interpretation ability. Compared with the deep learning methods, traditional methods can get more interpretable and stable results with model-based features. In order to take full advantage of these two kinds of methods, we propose target part attention network based on the attributed scattering center (ASC) model to integrate the electromagnetic characteristics with the deep learning framework. Firstly, considering the importance of scattering structure for SAR ATR, we design a target part model based on ASC model. Then, a novel part attention module based on Scaled Dot-Product Attention mechanism is proposed, which directly associates the features of target parts with the classification results. Finally, we give the derivation method of the importance of each part, which is of great significance for practical application and the interpretation of SAR ATR. Experiments on the MSTAR data set demonstrate the effectiveness of the proposed part attention network. Compared with existing studies, it can achieve higher and more robust classification accuracy under different complex conditions. Furthermore, combined with the importance of parts, we constructed two effective interpretable analysis methods for deep learning network classification results.

Recently, deep learning has made significant progress in synthetic aperture radar automatic target recognition (SAR ATR). However, deep convolutional neural networks (DCNNs) are discovered to be susceptible to carefully crafted adversarial perturbations. Regarding the unique scattering mechanism in SAR imaging, the scattering feature such as attributed scattering centers (ASCs) should be deeply considered in the adversarial attack (AA) algorithms for DCNNs in SAR ATR. In this letter, an AA algorithm named ASC-STA is proposed to take advantage of the powerful SAR imaging property characterization capability of ASC model. Considering the imaging characteristic which is presented in the ASC model, the spatial transformation module is specially designed to be performed on the strong back-scattering structures guided by the reconstruction image of ASC model. Spatial transformation shifts the robust scattering point features via altering the locally gray-scale relationships of the target microstructures in SAR images. Besides, a modified shape context evaluated metric is established to assess the validity of the feature alternation process. Experimental results on the MSTAR database demonstrate that the proposed method achieves a high success rate without complex parameter settings and generates the perturbed image in high image quality. Compared with the norm-based AA algorithms, which are not involved with ASCs, the point feature of the original image is efficiently altered. At the same time, the perturbations are focused on the target regions accurately.

Deep learning has been widely used in automatic target recognition (ATR) for synthetic aperture radar (SAR) recently. However, most of the studies are based on the network structure in optical images and lack full consideration of the inherent characteristics of SAR targets, which limits the improvement of recognition accuracy and makes poor generalization ability. In addition, due to the black-box characteristics, it is difficult to effectively interpret SAR ATR results. To conquer these problems, we propose an electromagnetic scattering feature (ESF) module embedded network based on attributed scattering center (ASC) model to incorporate the SAR targets’ characteristics into the deep learning framework. First, a novel convolutional neural network (CNN)-based algorithm for extracting ASC parameters is proposed, which makes the network focus on target features under the guidance of physical model. Then, the ESF module is designed based on a well-trained ASC parameters extractor to inject the learned target features into the classification network for more robust and interpretable results. Besides, two structures are proposed combined with the ESF module for single-view and multiview SAR target classification, which further illustrates the portability of the module. Experiments on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show the validity of the proposed CNN-based ASC extractor and the ESF module embedded classification network. Compared with ordinary networks, our method can achieve higher classification accuracy under complex conditions, which reflects the better generalization performance of the algorithm. Furthermore, through visualization analysis of the classification results, we show the interpretability of the network combined with the electromagnetic scattering characteristics.

Most deep-learning-based target detection methods have high computational complexity and memory consumption, and they are difficult to deploy on edge devices with limited computing resources and memory. To tackle this problem, this article proposes to learn a lightweight detector named Light-YOLOv4, and it is obtained from YOLOv4 through model compression. To this end, first, we perform sparsity training by applying L1 regularization to the channel scaling factors, so the less important channels and layers can be found. Second, channel pruning and layer pruning are enforced on the network to prune the less important parts, which could significantly reduce network's width and depth. Third, the pruned model is retrained with a knowledge distillation method to improve the detection accuracy. Fourth, the model is quantized from FP32 to FP16, and it could further accelerate the model with almost no loss of detection accuracy. Finally, to evaluate Light-YOLOv4's performance on edge devices, Light-YOLOv4 is deployed on NVIDIA Jetson TX2. Experiments on the SAR ship detection dataset (SSDD) demonstrate that the model size, parameter size, and FLOPs of Light-YOLOv4 have been reduced by 98.63%, 98.66%, and 91.30% compared with YOLOv4, and the detection speed has been increased to 4.2×. While the detection accuracy of Light-YOLOv4 is only slightly reduced, for example, the mAP has only reduced by 0.013. Besides, experiments on the Gaofen Airplane dataset also prove the feasibility of Light-YOLOv4. Moreover, compared with other state-of-the-art methods, such as SSD and FPN, Light-YOLOv4 is more suitable for edge devices.

Synthetic aperture radar (SAR) target recognition faces the challenge that there are very little labeled data. Although few-shot learning methods are developed to extract more information from a small amount of labeled data to avoid overfitting problems, recent few-shot or limited-data SAR target recognition algorithms overlook the unique SAR imaging mechanism. Domain knowledge-powered two-stream deep network (DKTS-N) is proposed in this study, which incorporates SAR domain knowledge related to the azimuth angle, the amplitude, and the phase data of vehicles, making it a pioneering work in few-shot SAR vehicle recognition. The two-stream deep network, extracting the features of the entire image and image patches, is proposed for more effective use of the SAR domain knowledge. To measure the structural information distance between the global and local features of vehicles, the deep Earth mover’s distance is improved to cope with the features from a two-stream deep network. Considering the sensitivity of the azimuth angle in SAR vehicle recognition, the nearest neighbor classifier replaces the structured fully connected layer for $K$ -shot classification. All experiments are conducted under the configuration that the SARSIM and the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset work as a source and target task, respectively. Our proposed DKTS-N achieved 49.26% and 96.15% under ten-way one-shot and ten-way 25-shot, whose labeled samples are randomly selected from the training set. In standard operating condition (SOC) as well as three extended operating conditions (EOCs), DKTS-N demonstrated overwhelming advantages in accuracy and time consumption compared with other few-shot learning methods in $K$ -shot recognition tasks.