SAR Target Detection Network via Semi-supervised Learning

The existing Synthetic Aperture Radar (SAR) image target detection methods based on convolutional neural networks (CNNs) have achieved remarkable performance, but these methods require a large number of target-level labeled training samples to train the network. Moreover, some clutter is very similar to targets in SAR images with complex scenes, making the target detection task very difficult. Therefore, a SAR target detection network based on a semi-supervised learning and attention mechanism is proposed in this paper. Since the image-level label simply marks whether the image contains the target of interest or not, which is easier to be labeled than the target-level label, the proposed method uses a small number of target-level labeled training samples and a large number of image-level labeled training samples to train the network with a semi-supervised learning algorithm. The proposed network consists of a detection branch and a scene recognition branch with a feature extraction module and an attention module shared between these two branches. The feature extraction module can extract the deep features of the input SAR images, and the attention module can guide the network to focus on the target of interest while suppressing the clutter. During the semi-supervised learning process, the target-level labeled training samples will pass through the detection branch, while the image-level labeled training samples will pass through the scene recognition branch. During the test process, considering the help of global scene information in SAR images for detection, a novel coarse-to-fine detection procedure is proposed. After the coarse scene recognition determining whether the input SAR image contains the target of interest or not, the fine target detection is performed on the image that may contain the target. The experimental results based on the measured SAR dataset demonstrate that the proposed method can achieve better performance than the existing methods.

As an active microwave imaging sensor, Synthetic Aperture Radar (SAR) has become one of the main means of Earth observation owing to its unique technical advantages of all-day, all-weather operation and long working distance. As such, it plays a very important role in military and civilian fields. With the development of SAR remote-sensing technology, high-resolution, high-quality SAR images are produced continuously. However, manual detection and recognition of targets of interest is time-consuming and laborious, so the development of Automatic Target Recognition (ATR) technology is a matter of urgency. The typical SAR ATR system primarily comprises three stages: detection, discrimination, and classification/recognition. The detection and discrimination stages are the basis of the SAR ATR system, and research on SAR applications in the radar field has been conducted by researchers around the world. For single-channel SAR images, target detection and discrimination from simple scenes yield good results. However, in complex scenes, the clutter scattering intensity is relatively high, the clutter background is heterogenous, the target scattering intensity is relatively weak, and the target distribution is dense. These factors continue to make accurate SAR target detection and discrimination difficult. In this paper, we summarize the recent research progress on single-channel SAR target detection and discrimination methods for complex scenes, analyze the characteristics and problems associated with various methods, and consider the future development trend of single-channel SAR target detection and discrimination methods for complex scenes.

The performance of deep learning approaches for Synthetic Aperture Radar (SAR) target detection is often limited by the scarcity of annotated data. While Self-Supervised Learning (SSL) has emerged as a powerful paradigm to mitigate data dependence, its potential in SAR target detection remains largely underexplored. In this study, we propose SARDet-MIM, a comprehensive framework based on Masked Image Modeling (MIM), to enhance SAR target detection. The approach consists of two stages. In the self-supervised pre-training stage, we propose an innovative Structural and Scattering Masked Autoencoder (SSMAE) method for SAR imagery. Unlike conventional MIM methods, which typically reconstruct raw pixels, SSMAE employs a physics-aware reconstruction target comprising multi-scale gradient and SAR-Harris features. This strategy explicitly guides the network to capture discriminative structural contexts and intrinsic scattering features that benefit SAR target detection. For downstream detection, we construct a Maximally Pre-trained Detector (MPD), which integrally transfers the pre-trained ViT encoder–decoder architecture to the detection network to fully exploit pre-trained representations. Extensive experiments on three SAR target detection datasets demonstrate that SARDet-MIM consistently outperforms competing methods.

Deep convolutional neural networks (DCNNs) have attracted extensive attention in synthetic aperture radar (SAR) image target detection. However, there is usually a lack of labeled data in SAR images which results in difficulties to improve the detection performance with DCNN based methods. Transfer learning borrows generic knowledge from other tasks for the target task and can be an effective solution to the small sample problem, but the features' generality and specificity across various models are not discussed in target detection. In addition, conventional DCNN based models have too many parameters related to the prior information of the targets which increase the difficulty of transfer learning between different tasks. In this paper, we mainly conduct research in three folds to address these mentioned issues. First, the generality and specificity of features in different network models are analyzed in the field of SAR target detection. Second, a transitive transfer framework is developed for detection under small sample condition. Third, an anchor free rotatable detector with flexible structure is designed for SAR images detection to suit the transfer detection framework proposed in this paper and a reasonable explanation is given for its upper hand with small samples. Experimental results demonstrate that: 1) transitive transfer framework is valid in SAR target detection; 2) the anchor free detector in this framework is superior to the anchor based one; 3) the proposed detector achieves better performance under small sample condition by comparison with state-of-the-art methods.

It is expensive and time-consuming to obtain a large number of labeled synthetic aperture radar (SAR) images. In the task of small training data size, the results of target detection on SAR images using deep network approaches are usually not ideal. In this study, considering that optical remote sensing images are much easier to be labeled than SAR images, we assume to have a large number of labeled optical remote sensing images and a small number of labeled SAR images with the similar scenes, propose to transfer knowledge from optical remote sensing images to SAR images, and develop a domain adaptive Faster R-CNN for SAR target detection with small training data size. In the proposed method, in order to make full use of the label information and realize more accurate domain adaptation knowledge transfer, an instance level domain adaptation constraint is used rather than feature level domain adaptation constraint. Specifically, generative adversarial network (GAN) constraint is applied as the domain adaptation constraint in the adaptation module after the proposals of Faster R-CNN to achieve instance level domain adaptation and learn the transferable features. The experimental results on the measured SAR image dataset show that the proposed method has higher detection accuracy in the task of SAR target detection with small training data size than the traditional Faster R-CNN.

A target detection method based on improved single shot multibox detector (SSD) is proposed to solve insufficient training samples for synthetic aperture radar (SAR) target detection. Firstly, a residual network is introduced in the feature extraction structure for extracting deep features of the target; then, the aspect ratio of default boxes is redesigned to match different sizes of target detection; finally, the existing dataset is expanded with small samples based on saliency map. The experimental results based on the simulated SAR data and the real SAR data prove that the proposed method has better performance in SAR target detection.

The deep learning object detection algorithm has been widely applied in the field of synthetic aperture radar (SAR). By utilizing deep convolutional neural networks (CNNs) and other techniques, these algorithms can effectively identify and locate targets in SAR images, thereby improving the accuracy and efficiency of detection. In recent years, achieving real-time monitoring of regions has become a pressing need, leading to the direct completion of real-time SAR image target detection on airborne or satellite-borne real-time processing platforms. However, current GPU-based real-time processing platforms struggle to meet the power consumption requirements of airborne or satellite applications. To address this issue, a low-power, low-latency deep learning SAR object detection algorithm accelerator was designed in this study to enable real-time target detection on airborne and satellite SAR platforms. This accelerator proposes a Process Engine (PE) suitable for multidimensional convolution parallel computing, making full use of Field-Programmable Gate Array (FPGA) computing resources to reduce convolution computing time. Furthermore, a unique memory arrangement design based on this PE aims to enhance memory read/write efficiency while applying dataflow patterns suitable for FPGA computing to the accelerator to reduce computation latency. Our experimental results demonstrate that deploying the SAR object detection algorithm based on Yolov5s on this accelerator design, mounted on a Virtex 7 690t chip, consumes only 7 watts of dynamic power, achieving the capability to detect 52.19 512 × 512-sized SAR images per second.

This paper studies the polarimetric synthetic aperture radar (SAR) image interpretation and target detection. The polarimetric target decomposition theorem and dominant scattering mechanism identification approach have been used widely in polarimetric SAR image interpretation. In order to get a good performance, we present a three-step image interpretation approach which involves speckle reduction and image classification, as pre-processing before the classical interpretation approach. Furthermore, we study the target detection based on the covariance matrix decomposition, and an odd-scattering detector and an even-scattering detector are formed and analyzed. Experimental results are given to verify the effectiveness of the image interpretation approach and to quantify the presented polarimetric detectors.

Synthetic-aperture radar (SAR) image target detection is widely used in military, civilian and other fields. However, existing detection methods have low accuracy due to the limitations presented by the strong scattering of SAR image targets, unclear edge contour information, multiple scales, strong sparseness, background interference, and other characteristics. In response, for SAR target detection tasks, this paper combines the global contextual information perception of transformers and the local feature representation capabilities of convolutional neural networks (CNNs) to innovatively propose a visual transformer framework based on contextual joint-representation learning, referred to as CRTransSar. First, this paper introduces the latest Swin Transformer as the basic architecture. Next, it introduces the CNN’s local information capture and presents the design of a backbone, called CRbackbone, based on contextual joint representation learning, to extract richer contextual feature information while strengthening SAR target feature attributes. Furthermore, the design of a new cross-resolution attention-enhancement neck, called CAENeck, is presented to enhance the characterizability of multiscale SAR targets. The mAP of our method on the SSDD dataset attains 97.0% accuracy, reaching state-of-the-art levels. In addition, based on the HISEA-1 commercial SAR satellite, which has been launched into orbit and in whose development our research group participated, we released a larger-scale SAR multiclass target detection dataset, called SMCDD, which verifies the effectiveness of our method.

Synthetic Aperture Radar (SAR) images are prone to be contaminated by noise, which makes it very difficult to perform target recognition in SAR images. Inspired by great success of very deep convolutional neural networks (CNNs), this paper proposes a robust feature extraction method for SAR image target classification by adaptively fusing effective features from different CNN layers. First, YOLOv4 network is fine-tuned to detect the targets from the respective MF SAR target images. Second, a very deep CNN is trained from scratch on the moving and stationary target acquisition and recognition (MSTAR) database by using small filters throughout the whole net to reduce the speckle noise. Besides, using small-size convolution filters decreases the number of parameters in each layer and, therefore, reduces computation cost as the CNN goes deeper. The resulting CNN model is capable of extracting very deep features from the target images without performing any noise filtering or pre-processing techniques. Third, our approach proposes to use the multi-canonical correlation analysis (MCCA) to adaptively learn CNN features from different layers such that the resulting representations are highly linearly correlated and therefore can achieve better classification accuracy even if a simple linear support vector machine is used. Experimental results on the MSTAR dataset demonstrate that the proposed method outperforms the state-of-the-art methods.

In recent years, convolutional neural network (CNN)-based methods have been extensively explored for synthetic aperture radar (SAR) target detection. Nevertheless, the convolutional sampling locations of CNNs cannot accurately fit vehicle targets due to the fixed sampling mechanism in the convolutional kernel. In this paper, we focus on the vehicle target detection task in SAR images and propose a novel rectangle-invariant rotatable convolution (RIRConv) to determine more accurately the convolutional sampling locations for vehicle targets. Specifically, this paper considers the shape characteristic of vehicle targets in SAR images, which always retain a rectangular shape despite having varying sizes, aspect ratios, and rotation angles. The proposed RIRConv equips three additional learnable attribute parameters, namely, width, height, and angle attributes, to adaptively adjust the sampling locations in the convolutional kernel according to the targets. In addition, the RIRConv applies a modulation mechanism to focus on the sampling locations that significantly affect the output. Finally, the RIRConv is introduced into the single-shot multibox detector (SSD) to realize SAR vehicle target detection. In this way, the feature representation capability of SSD for vehicle targets can be enhanced, thus leading to higher detection performance. Notably, the proposed RIRConv is “plug-and-play” and can also be used with other existing advanced technologies to achieve higher detection performance. The experiments based on the measured miniSAR data validate the effectiveness of the proposed method.

This tutorial presents a state-of-the-art survey of advanced applications of Synthetic Aperture Radar (SAR). Applications include: the detection and recognition of targets in SAR imagery, and the detection of targets under trees (and in the open) using change detection (coherent change detection, full-polarization change detection, etc). A fully polarimetric SAR sensor is described. Models of clutter and targets are developed from analysis of full-polarization SAR imagery. Algorithms for optimum (GLRT) processing of full-polarization data are derived, including the polarimetric whitening filter (PWF) and the polarimetric matched filter (PMF). Polarization-based discrimination features (polarization entropy, anisotropy, etc.) are developed and applied to SAR image terrain classification. Model-based and template-based target recognition systems are described and classifier performance results versus resolution and polarization are summarized for 10-and 20-target classifiers. Super-resolution algorithms are applied to SAR imagery and target recognition performance improvements demonstrated. Multi-polarization and multi-pass change detection algorithms are developed and demonstrated using UHF SAR imagery of targets under trees; fusion of multi-polarization coherent and non-coherent change detection algorithms is demonstrated. The effect of SAR image compression on change detection performance is quantified. Change detection performance using multi-pass VHF SAR imagery of targets under trees is demonstrated using data gathered by the Swedish CARABAS SAR system.

Synthetic Aperture Radar (SAR) target classification is an important branch of SAR image interpretation. The deep learning based SAR target classification algorithms have made remarkable achievements. But the acquisition and annotation of SAR target images are time-consuming and laborious, and it is difficult to obtain sufficient training data in many cases. The insufficient training data can make deep learning based models suffering from over-fitting, which will severely limit their wide application in SAR target classification. Motivated by the above problem, this paper employs transfer-learning to transfer the prior knowledge learned from a simulated SAR dataset to a real SAR dataset. To overcome the sample restriction problem caused by the poor feature discriminability for real SAR data. A simple and effective sample spectral regularization method is proposed, which can regularize the singular values of each SAR image feature to improve the feature discriminability. Based on the proposed regularization method, we design a transfer-learning pipeline to leverage the simulated SAR data as well as acquire better feature discriminability. The experimental results indicate that the proposed method is feasible for the sample restriction problem in SAR target classification. Furthermore, the proposed method can improve the classification accuracy when relatively sufficient training data is available, and it can be plugged into any convolutional neural network (CNN) based SAR classification models.

Ship detection from synthetic aperture radar (SAR) images is a hot topic, but the difficulty in collecting labeled SAR images may hinder the development of deep-learning-based detection methods. Inspired by the idea of domain adaptation, in this article, we propose a hierarchical similarity alignment neural network (HSANet) for ship detection in SAR images, which is a domain adaptive (DA) approach with optical remote sensing images as training samples. The kernel target of HSANet is to mine and align both the global structure and the local instance information between SAR and optical images, where two modules, structural alignment module (SAM) and prototype alignment module (PAM), are designed to, respectively, conduct two hierarchies of alignment process. In general, SAM attempts to extract the global structure similarity which exists in image-level feature representation, while PAM tends to extract the local shape similarity which is instance-level representation. To be specific, SAM is developed by Fourier-based feature alignment, which tries to describe the similar structural relationship between optical and SAR images. Meanwhile, PAM is proposed based on the conjoint confidence analysis where the instance-level ship representations of the source and target domains is aligned. SAM and PAM work together to construct a hierarchical domain adaptation network for SAR ship detection. Experiments on several public datasets may indicate the effectiveness of the proposed method.

Thanks to the availability of large-scale data, deep Convolutional Neural Networks (CNNs) have witnessed success in various applications of computer vision. However, the performance of CNNs on Synthetic Aperture Radar (SAR) image classification is unsatisfactory due to the lack of well-labeled SAR data, as well as the differences in imaging mechanisms between SAR images and optical images. Therefore, this paper addresses the problem of SAR image classification by employing the Generative Adversarial Network (GAN) to produce more labeled SAR data. We propose special GANs for generating SAR images to be used in the training process. First, we incorporate the quadratic operation into the GAN, extending the convolution to make the discriminator better represent the SAR data; second, the statistical characteristics of SAR images are integrated into the GAN to make its value function more reasonable; finally, two types of parallel connected GANs are designed, one of which we call PWGAN, combining the Deep Convolutional GAN (DCGAN) and Wasserstein GAN with Gradient Penalty (WGAN-GP) together in the structure, and the other, which we call CNN-PGAN, applying a pre-trained CNN as a discriminator to the parallel GAN. Both PWGAN and CNN-PGAN consist of a number of discriminators and generators according to the number of target categories. Experimental results on the TerraSAR-X single polarization dataset demonstrate the effectiveness of the proposed method.