CRTransSar: A Visual Transformer Based on Contextual Joint Representation Learning for SAR Ship Detection

Abstract

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.

Feature Enhancement Pyramid and Shallow Feature Reconstruction Network for SAR Ship Detection

“…After the recent appearance of the popular visual transformer (ViT), some transformer models have been introduced into SAR ship detection, considering the advantages of ViT in establishing long-range dependencies. Xia et al [36] proposed a ViT architecture, called CRTransSar, to enhance context learning by combining transformer and CNN. Li et al [37] introduced the Swin transformer as the backbone in cascade R-CNN to improve feature extraction ability and proposed a feature fusion module to optimize the feature fusion capability of the feature pyramid.…”

Feasibility Study on the Emergency Acquisition of Satellite Images for Maritime Search and Rescue

“…To minimize damage from maritime accidents, a control system that can urgently grasp the current situation and respond to accidents is needed. To this end, remote sensing platforms such as drones, helicopters, manned aircraft and satellites have attracted considerable attention as surveillance means and applied in various scenarios such as maritime debris detection (Choi, 2021), ship detection (Xia et al, 2022), and oil spill monitoring (Zhang et al, 2022). Drones, helicopters and manned aircraft can perform precise detection with high spatial resolution, but the coverage area is narrow as they fly at low altitude.…”

Grid-guided localization network based on the spatial attention mechanism for synthetic aperture radar ship detection

“…Deep learning has the advantages of high precision, speed, and ability to perform end-to-end target detection, and deep learning-based networks have been widely used in the field of SAR ship detection in recent years. Xia et al 13 proposed the CRTransSar, which is a visual transformer framework based on contextual joint-representation learning. While enhancing the SAR target feature attributes, the CRTransSar can extract richer context feature information, enabling high accuracy, but its speed is limited.…”