Small ship detection in SAR images based on modified SSD
Long Han, Wei Ye, Jianwei Li, Da Ran · 2019 IEEE International Conference on Signal, Information and Data Processing (ICSIDP) · 2019
Small object detection is a nodus for CNNs (Convolutional Neural Networks) that take advantages of feature pyramid for object detection, and it is also a hotspot which researchers focus on, especially small ship detection in SAR images. Based on SSD (Single Shot Multibox Detector), which is a classical CNN detector and a typical one-stage method for object detection, and inspired by the ideas of object enhancement and feature fusion, the paper uses deconvolution to enhance representation of small object in feature pyramid. Thereby we expect to improve detecting precision of SSD. The feasibility and effectiveness of proposed method are proved by our experiments. Experimental results on SSDD show that the proposed method greatly improves detecting precision of small ships in SAR images.