An Improved YOLO-v3Algorithm for Ship Detection in SAR Image Based on K-means++ with Focal Loss

In recent years, practical industrial production application have put forward extremely high requirements for its detection accuracy and detection efficiency in the aspect of synthetic aperture radar (SAR) image ship detection. Among the solutions to this problem, the YOLO has received more and more attention due to its advantages such as high speed. In this paper, the K-means++ is used to obtain the Anchor Box, the Focal loss is introduced to balance the proportion of positive and negative samples, and an improved image detection algorithm based on YOLO-v3 is proposed to solve the low detection efficiency and detection accuracy of ship images. The experimental results show that the improved algorithm in this paper can get rid of the local optimum well, shorten the convergence time, improve the training efficiency and the accuracy of ship image detection.