Robust single stage detector based on two-stage regression for SAR ship detection
Shiqi Chen, Ronghui Zhan, Jun Zhang · 2018
Automatic ship detection in SAR imagery plays an indispensable role in the surveillance of maritime activity. With the spring of the deep neural network and the rapid development of SAR imaging technique, SAR ship detection based on deep convolutional neural networks has attracted increasing attention in remote sensing imagery interpretation. However, the ships of variant sizes (multi-scale) and complicated background result in the essential challenge of using deep learning based methods. The initial motivation of this paper is to design a structure that can achieve better accuracy as well as maintain compatible time efficiency on small object detection. Aiming at solving this problem, we develop anelaborately designed network based on single-shot detector and two-stage regression called Robust two-stage regression network (R2RN). Our network comprises an anchor modified first module and an object detection second module as well as the connective block between them which inherits the essence of feature pyramid. Experiments on SAR ship dataset have demonstrated the effectiveness and efficiency of the proposed method.