Abstract:Aiming at the problems of low accuracy and easy to miss detection in traditional synthetic aperture radar ship image detection methods, this paper proposes an image ship detection method based on C3K2-SC-Yolov11. Firstly, the parallel convolutional multi-scale feature module C3K2_SC was designed to replace the C3K2 module, effectively enhancing the adaptive ability of the model, enabling it to simultaneously focus on information-rich channels and important regions, and capture salient targets in the global context. Then, the HSAN_PSA module was designed to replace C2PSA. By deeply integrating hierarchical spatial attention and position-sensitive attention, the interference of coherent speckle noise was significantly suppressed. Finally, the SSSA attention mechanism was introduced to improve CBAM. While maintaining the advantages of CBAM, it significantly enhanced its performance in small object detection and multi-scale adaptation, making it suitable for object detection tasks in complex scenarios. The experimental results show that the improved model achieves a 6% and 9.59% improvement in mAP@0.5 and mAP@50-95 on the HRSID dataset compared to YOLOv11, respectively.