Intelligent detection of ship wakes in ocean SAR images

Abstract

Ship wakes in Synthetic Aperture Radar (SAR) imagery demonstrate the coupling between ship and ocean environment, holding significant potential for ocean remote sensing applications. However, under complex sea states, wakes typically exhibit low signal-to-noise ratios, resulting in weak texture features that hinder reliable automatic detection. To address this challenge, this study develops an integrated simulation framework that incorporates dynamic sea surface evolution, ship wake generation, and SAR imaging, enabling the generation of ship wake samples under diverse sea conditions and radar imaging configurations. Furthermore, a high-precision detection model tailored for weak wakes, termed YOLOv11- DAWA, is constructed by incorporating a Direction-Aware Weighted Attention (DAWA) module. The key innovation lies in the wake-oriented design of DAWA and its embedding into YOLOv11 to enhance the representation of directional and weak wake features. Experimental results demonstrate that the proposed model achieves robust detection performance under complex sea conditions and weak wake scenarios. Compared with the baseline YOLOv11, the proposed model improves precision, recall, mAP@0.5, mAP@0.5:0.95, and F1-score by 5.35%, 7.81%, 5.02%, 7.23%, and 6.74%, respectively. Overall, this work establishes an end-to-end framework from physical simulation to intelligent detection, providing a robust solution for high-precision weak wake detection under complex sea states.