SAR image targets detection and recognition (SAR-TDR) have become research hotpots in the remote sensing application. These targets include ships, vehicles, aircrafts, oil tanks, bridges, and so on. However, with the rapid development of SAR technology and increasingly complex electromagnetic environment, complex characteristics of SAR images bring severe challenges to the accurate SAR-TDR via traditional physical models or manually features extraction based machine learning methods. In recent years, deep learning (DL), as a powerful automatic features extraction algorithm, has been widely used in the computer vision domain. More specifically, DL has also been introduced into the SAR-TDR tasks and effectively achieved good performances in terms of accuracy, real-time, etc. With the rapid development of DL, SAR images processing, and practical requirements of SAR-TDR in civilian and military domains, it is very necessary to make a systematic survey on the SAR-TDR in past few years. In this survey paper, we mainly conduct a systematic overview of DL-based SAR-TDR literatures on two tasks, i.e., targets recognition (e.g., ground vehicles, ships, and aircrafts) and targets detection (e.g., ships, aircrafts,, change detection, seasurface oil spill, and oil tanks), respectively. More specifically, our related works about these topics are also presented to verify the effectiveness of DL-based methods. Firstly, several DL methods (e.g., CNN, RNN, AEs, and GNN), commonly used to SAR-TDR, are briefly introduced. The systematic review of DL-based SAR-TDR (including our related works) is then presented. Finally, the current challenges and future possible research directions are deeply analyzed and discussed.

Radar signal sorting (RSS) plays an important role in the electronic support measurement system. However, the existing clustering-based RSS methods depend heavily on prior knowledge to achieve excellent performance, which may bring severe challenges to RSS in actual scenarios. This letter proposes a novel subspace decomposition based adaptive density peak clustering (SD-ADPC) method to address the problems of low accuracy and high computational cost in RSS. First, the original complex radar signal data is directly decomposed into two-dimensional (2D) subspace by t-distributed random neighborhood embedding (t-SNE). Then, based on the outlier detection of the products between the peak density and the distance of the data points, ADPC is used to adaptively determine the optimal clustering centers of the original data in 2D subspace. Finally, the reminding data is assigned to its nearest cluster with Euclidean distance in one step. The experimental results of the simulated RSS dataset and the open baselines show that our proposed method does not require any knowledge and can achieve better or competitive performance in terms of accuracy and computational cost, compared to existing state-of-the-art methods.

Deep-learning-based synthetic aperture radar automatic target recognition (SAR-ATR) plays a significant role in the military and civilian fields. However, data limitation and large computational cost are still severe challenges in the actual application of SAR-ATR. To improve the performance of the convolutional neural network (CNN) model with limited data samples in SAR-ATR, this article proposes a novel multidomain feature subspace fusion representation learning method, i.e., a lightweight cascaded multidomain attention network, namely, LW-CMDANet. First, we design a four-layer CNN model to perform hierarchical feature representation learning via the hinge loss function, which can efficiently alleviate the overfitting problem of the CNN model by a nongreedy training style with a small dataset. Then, a cascaded multidomain attention module, based on discrete cosine transform and discrete wavelet transform, is embedded into the previous CNN to further complete the class-specific feature extraction from both the frequency and wavelet transform domains of the input feature maps. Thus, the multidomain attention can enhance the feature extraction ability of previous nongreedy learning manner, to effectively improve the recognition accuracy of the CNN model. Experimental results on small SAR datasets show that our proposed method can achieve better or competitive performance than that of many current existing state-of-the-art methods in terms of recognition accuracy and computational cost.

The dense, complex and variable electromagnetic environment poses a serious challenge to radar signal sorting (RSS) in modern electronic reconnaissance systems. In order to improve RSS performance, this letter proposes a semi-supervised learning framework-based RSS method via a residual graph convolutional network (ResGCN-RSS) to effectively improve the generalization ability of the signal sorting models in small data scenarios. Firstly, the graph structure construction of intercepted radar signals is performed via K-nearest neighbor algorithm. Then, the three-layer ResGCN is designed to adaptively improve the features learning. Finally, RSS can be effectively and efficiently implemented through an end-to-end ResGCN with small labeled graph data of interleaved radar signals. The simulation experimental results show that our proposed method can achieve better average accuracy with little computational cost increasing when the labeled data is very small, compared to some existing methods.

Intelligent ship detection based on high-precision synthetic aperture radar (SAR) images plays a vital role in ocean monitoring and maritime management. Denoising is an effective preprocessing step for target detection. Morphological network-based denoising can effectively remove speckle noise, while the smoothing effect of which blurs the edges of the image and reduces the detection accuracy. The fusion of edge extraction and morphological network can improve detection accuracy by compensating for the lack of edge information caused by smoothing. This article proposes an end-to-end lightweight network called morphological feature-pyramid Yolo v4-tiny for SAR ship detection. First, a morphological network is introduced to preprocess the SAR images for speckle noise suppression and edge enhancement, providing spatial high-frequency information for target detection. Then, the original and preprocessed images are combined into the multichannel as an input for the convolution layer of the network. The feature pyramid fusion structure is used to extract the high-level semantic features and shallow detailed features from the image, improving the performance of multiscale target detection. Experiments on the public SAR ship detection dataset and AIR SARShip-1.0 show that the proposed method performs better than the other convolution neural network-based methods.

In order to more efficiently mine the features of PolSAR images and build a more suitable classification model that combines the features of the polarimetric domain and the spatial domain, this paper proposes a PolSAR image classification method, called low-frequency and contour sub-bands driven polarimetric Squeeze-and-Excitation network (LC-PSENet). First, the proposed LC-PSENet introduces the nonsubsampled Laplacian pyramid to decompose polarimetric feature maps, so as to construct a multi-channel PolSAR image based on the low-frequency sub-band and contour sub-band of these maps. It guides the network to perform feature mining and selection in the sub-bands of each polarimetric map in a supervised way, automatically balancing the contributions of polarimetric features and their sub-bands and the influence of interference information such as noise, making the network learning more efficient. Second, the method introduces Squeeze-and-Excitation operation in CNN to perform channel modeling on the polarimetric feature sub-bands. It strengthens the learning of the contributions of local maps of the polarimetric features and sub-bands, thereby effectively combining the features of the polarimetric domain and the spatial domain. Experiments on the datasets of Flevoland, Netherlands and Oberpfaffenhofen show that the proposed LC-PSENet achieves overall accuracies of 99.66%, 99.72%, and 95.89%, which are 0.87%, 0.27% and 1.42% higher than the baseline CNN, respectively. The isolated points in the classification results are obviously reduced, and the distinction between boundary and non-boundary is more clear and delicate. Also, the method performs better than many current state-of-the-art methods in terms of classification accuracy.