Deep learning contributes to significant improvements in synthetic aperture radar (SAR) target recognition performance. Most SAR target recognition methods are based on supervised learning and require labeled SAR data. There only exists limited labeled data due to the time-consuming and laborious work of labeling, and there is still a large amount of available unlabeled radar data. Therefore, we aim to explore whether unlabeled data can provide the network with sufficient feature information and enable the network to cluster similar target features and distinguish different target features, thereby improving the SAR target recognition performance. In this paper, we propose a new framework to train a deep neural network for SAR target recognition to eliminate the need for a large amount of labeled training data. Our idea is based on transferring knowledge from a self-supervised model, where the data can train without label information. Experiments are performed on the moving and stationary target acquisition and recognition (MSTAR) benchmark dataset, and the experimental results demonstrate the improvements in recognition performance achieved by our proposed method with limited labeled data.

Existing synthetic aperture radar automatic target recognition (SAR ATR) methods have been effective for the classification of seen target classes. However, it is more meaningful and challenging to distinguish the unseen target classes, i.e., open set recognition (OSR) problem, which is an urgent problem for the practical SAR ATR. The key solution of OSR is to effectively establish the exclusiveness of feature distribution of known classes. In this letter, we propose an entropy-awareness meta-learning method that improves the exclusiveness of feature distribution of known classes which means our method is effective for not only classifying the seen classes but also encountering the unseen other classes. Through meta-learning tasks, the proposed method learns to construct a feature space of the dynamic-assigned known classes. This feature space is required by the tasks to reject all other classes not belonging to the known classes. At the same time, the proposed entropy-awareness loss helps the model to enhance the feature space with effective and robust discrimination between the known and unknown classes. Therefore, our method can construct a dynamic feature space with discrimination between the known and unknown classes to simultaneously classify the dynamic-assigned known classes and reject the unknown classes. Experiments conducted on the moving and stationary target acquisition and recognition (MSTAR) dataset have shown the effectiveness of our method for SAR OSR.

Convolutional neural networks (CNNs) have dominated the synthetic aperture radar (SAR) automatic target recognition (ATR) for years. However, under the limited SAR images, the width and depth of the CNN-based models are limited, and the widening of the received field for global features in images is hindered, which finally leads to the low performance of recognition. To address these challenges, we propose a Convolutional Transformer (ConvT) for SAR ATR few-shot learning (FSL). The proposed method focuses on constructing a hierarchical feature representation and capturing global dependencies of local features in each layer, named global in local. A novel hybrid loss is proposed to interpret the few SAR images in the forms of recognition labels and contrastive image pairs, construct abundant anchor-positive and anchor-negative image pairs in one batch and provide sufficient loss for the optimization of the ConvT to overcome the few sample effect. An auto augmentation is proposed to enhance and enrich the diversity and amount of the few training samples to explore the hidden feature in a few SAR images and avoid the over-fitting in SAR ATR FSL. Experiments conducted on the Moving and Stationary Target Acquisition and Recognition dataset (MSTAR) have shown the effectiveness of our proposed ConvT for SAR ATR FSL. Different from existing SAR ATR FSL methods employing additional training datasets, our method achieved pioneering performance without other SAR target images in training.

Synthetic aperture radar (SAR) automatic target recognition (ATR) is an essential field in SAR application. However, a sufficient number of labeled training SAR images for each target type plays a crucial role in existing SAR ATR methods, while the acquisition and annotation of SAR images are difficult and time-consuming in practice. Therefore, the recognition under the limited labeled training SAR images is the basic and crucial problem in SAR application. In this paper, we propose a novel hierarchically-designed lightweight method (HDLM) by recognition in label and discrimination in feature to address the problem of limited data in SAR ATR. The proposed method is hierarchically designed from top to bottom. In the top phase, the framework is constructed by dual loss to force the deep model to optimize by label recognition and feature discrimination, which is noted as recognition in label and discrimination in feature. In the middle phase, the architecture of the network is built up using a novel lightweight extractor and multi-level cross fusion to boost the amount and diversity of the features for the framework. In the bottom phase, two modules, coordinate attention, and depth-wise separable convolution modules are employed to enhance the feature quality and density with fewer parameters for the phases above. The experimental results on MSTAR and OpenSARship showed that the proposed HDLM performs better than the existing methods under the limited training samples.

Convolutional neural networks (CNN) show superior potential in synthetic aperture radar automatic target recognition (SAR ATR). However, due to the difficulty of obtaining SAR images and the scarcity of labeled SAR images, supervised learning has poor performance in this area and is not widely applicable. To address this problem, a semi-supervised conditional generative adversarial network with a multi-discriminator (SCGAN-MD) is proposed in this paper. In our method, a conditional generative adversarial network (CGAN) is adopted with two discriminators for training the generated images and predicting the labels for unlabeled samples. Compared with other semi-supervised learning-based methods, our proposed method has more accurate image generation capability and can achieve improved recognition accuracy of SAR ATR. Experiments on the Moving and Stationary Target Acquisition and Recognition (MSTAR) database indicate that the proposed method can effectively improve the recognition accuracy and robustness of the network with a small number of labeled samples.