AN MULTILAYER FUSION STRATEGY BASED ON IMPROVED YOLOV5 FOR SHIP DETECTION IN SAR IMAGES

Since entering the era of deep learning, the single-stage detection algorithm represented by YOLO has achieved some progress in the detection of ship targets by synthetic aperture radar (SAR). However, the accuracy of single-stage detection is lower than that of two-stage detection, especially for small target detection. To this end, this paper proposes an improved YOLOv5 detection method based on convolutional bolck attention module CBAM) and WPAN. At the same time, the focal efficient intersection over union (Focal EIoU) is introduced to optimize the calculation of bounding box regression loss. The experiment is verified on SAR ship detection dataset(SSDD). The results show that the proposed improved YOLOv5 algorithm can improve the false alarm and missing detection problems in multi-scale target detection, and improve the detection accuracy consequently.