Ship Target Detection Algorithm for SAR Images Based on YOLOv8-obb Rotating Bounding Box

With the development of rotating bounding box target detection technology, the use of rotating bounding box for ship detection in synthetic aperture radar (SAR) images has become a research hotspot. However, due to the small ship target, high speckle noise, side flap effect, unclear boundary and other reasons in SAR images, there are still some obstacles in rotating bounding box SAR ship image detection. To address the above problems, this paper proposes a rotating frame SAR image ship target detection algorithm based on YOLOv8-obb. First, using YOLOv8n-obb as the benchmark network, the SPPF_LSKA module fusing LSKA attention is proposed to enhance the feature extraction ability of the model for SAR ship images. Secondly, the noise suppression effect of the image is enhanced by introducing the DySample module. The robustness and reliability of the model are verified through a large number of experiments conducted on the standard datasets SSDD, HRSID and RSDDSAR. The experimental results show that compared with YOLOv8n-obb, the improved model improves the accuracy by 2.7% on SSDD and 1% on mAP50-95; by 1.1% on HRSID and 0.2% on mAP50-95; and by 1.1% on RSDD-SAR and 0.2% on mAP50-95. This model outperforms the existing SAR image ship detection model with higher accuracy.