SAR image ship detection based on improved YOLOv4
Qiwei Lin, Bowen Wang, Wang Yanfeng · 2021
Ship target detection is an important means to supervise and protect maritime rights and interests and is widely used in the marine monitoring field. To improve the detection accuracy of ship targets in SAR (Synthetics Aperture Radar) images, this paper proposes a SAR image ship target detection algorithm on the improved YOLOv4. The advantages of various improvement techniques of the YOLOv4 detection algorithm are analyzed, and the YOLOv4 improvement techniques that are most suitable for this data set are selected through experiments. Then adjust the anchor box of the original YOLOv4 algorithm based on the K-means clustering algorithm to solve the detection accuracy drop caused by the mismatch between the original anchor box and the ship target size ratio. Finally, in the open SSDD data sets for the improved algorithm was trained and tested. The experimental results show that, under the condition of ensuring the detection efficiency, the detection accuracy of the method in this paper is improved by 2.87% compared with the original YOLOv4 algorithm.