An Effective Ship Detection Method Based on RefineDet in SAR Images
Mingming Zhu, Guoping Hu, Shuai Li, Shiping Liu, Shiqiang Wang · 2021
Ship detection in synthetic aperture radar (SAR) imagery is a fundamental and challenging task. In view of the poor detection performance of traditional methods, we present an effective ship detection method based on RefineDet, which consists of three main parts, namely, an anchor refinement module, an object detection module and a transfer connection block. The proposed method can use the loss function to train uniformly. Experimental results on SSDD dataset illustrate that the proposed method obtains 98.4% AP, whose performance is significantly better than other methods, such as Faster R-CNN, SSD, RetinaNet, YOLOv2, and YOLOv3.