An Effective Method for Small Targets Detection in Synthetic Aperture Radar Images Under Complex Background

Abstract

Synthetic Aperture Radar (SAR) is a useful tool in marine surveillance. Small targets detection in SAR images especially in nearshore area is a difficult issue. Due to the complex background, there exist a lot of false targets. Therefore, we propose an effective method for small targets detection in SAR images under complex background, which combines the features of SAR images and those of SAR time series. A new neural network which integrates a neighborhood similarity module is constructed to enhance the features of small targets in SAR images. Then, a false alarm suppression method is put forward, which is based on empirical orthogonal functions to extract spatio-temporal features. Compared with other false alarm suppression methods, our proposed method is easily-implemented, highly efficient and in no need of a priori information. Simulation results on real datasets prove the efficiency and effectiveness of our proposed method.INDEX TERMS SAR detection, complex background, false alarm suppression, neural network.

A Fast and Accurate Small Target Detection Algorithm Based on Feature Fusion and Cross-Layer Connection Network for the SAR Images

“…Because of the principle of SAR imaging, some small targets which have few feature information usually exist in the images, and it's difficult to separate these small targets from the background correctly. To solve this problem, the algorithm combining the features of SAR image and time series [16], the algorithm constructing global attention modules in the spatial and channel domains [17], Lin et al [18] proposed a new network architecture based on the faster R-CNN to further improve the detection performance by using squeeze and excitation mechanism to improve the detection performance, Sun et al [19] proposed an anchor-free method for ship target detection in HR SAR images to address the complex surroundings, targets defocusing, and diversity of the scales, and obtain encouraging detection performance 2 > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DOUBLE-CLICK HERE TO EDIT) < compared other networks, Sun et al [20] proposed a novel YOLO-based arbitrary-oriented SAR ship detector using bidirectional feature fusion and angular classification (BiFA-YOLO) to address the multi-scale, arbitrary directions and dense arrangement issues, this method shows strong robustness and adaptability in HR SAR images, Kuang et al [21] proposed an elaborately designed deep hierarchical network based convolutional neural network with multilayer fusion to improve the detection performance for small-sized ships, Song et al [22] proposed an attention-guided end-to-end change detection network (AGCDetNet) based on the fully convolutional network and attention mechanism to the detection performance of high-resolution remote sensing images. The algorithm introducing spatial attention and channel attention mechanisms during the feature extracting stage [23], the algorithm introducing a rotating bounding boxbased target detection algorithm that can effectively reduce the interference of background pixels and avoid overlapping detection boxes of dense targets [24], and the algorithm embedding an enhanced attention module into the RCNN [25] are proposed.…”

Lightweight and anchor-free frame detection strategy based on improved CenterNet for multiscale ships in SAR images

“…With the rapid development of high-resolution synthetic aperture radar (SAR) imaging technology (Xie et al, 2016(Xie et al, , 2017(Xie et al, , 2020(Xie et al, , 2022, more SAR satellites have been launched, and along with it, the amount of SAR image data has grown rapidly. Subsequently, a large number of studies on target detection using SAR images have emerged (Chen et al, 2022;Tang et al, 2022;Xu et al, 2022), especially in the application of ocean monitoring (Cui et al, 2022;Kahar et al, 2022;Song et al, 2022). Generally, ship detection using SAR images can be divided into far-sea ship detection, near-sea (coastal) ship detection, and nearshore (port) ship detection (Li et al, 2022).…”