SAR ship detection based on multi-scale feature enhancement
Shun He, Yuzhu Wang, Zhiwei Yang · Remote Sensing Letters · 2025
In SAR images, ships are sparsely distributed and have various scales, which significantly impacts the accuracy of existing object detection; this paper introduces an improved object detection method for multi-scale SAR ship detection tasks. First, to improve the detection correctness of multi-scale SAR ships, the multi-scale feature extraction block (MFEB) was introduced. Second, to strengthen feature expression and improve the detection robustness in complex scenes, we use the deformable feature module (DFM). Finally, the extracted features are sent to the SR-Head to adaptively focus on crucial information and reduce missed detection and false detection problems. We conducted copious experiments on the two public datasets, SSDD and HRSID, and the results show that our detection model has good detection performance and high application value.