Rapid Ship Detection in SAR Images Based on YOLOv3

Abstract

With the increasing resolution and data volume of synthetic aperture radar images, ship detection in synthetic aperture radar images has become one of the hot spots of academic research. In recent years, object detection methods based on deep convolutional neural networks have gradually become the mainstream methods in the field of object detection based on natural images. To address the problems of low accuracy rate and detection speed of ship detection methods in synthetic aperture radar images, an end-to-end ship detection method based on YOLOv3 is proposed. Unlike the previous predicted position offset, we directly predict the position coordinates of the detection frame and set the parameters of the anchor frame through dimension clusters. The multi-scale output combines high-level semantic information from high-level feature maps and detailed information from low-level feature maps. The simulation results show that our proposed method is more accurate and faster than other methods.