Visual Context Aware Ship Detector for High-Resolution SAR Imagery

The ever increasing resolution of synthetic aperture radar (SAR) images have made accurate scene understanding possible. However, the rich visual context information in SAR images is not yet well explored for model construction, which limited the performance of traditional target detectors. In this paper, we propose a visual context aware approach to mimic the human visual perception mechanism for ship detection from high-resolution SAR imagery. Rather than considering target as a group of individual scatters, we adopt object-level selective search to efficiently locate sparse targets from wide SAR scenes. To verify the effectiveness of the proposed method, we test it along with the Hebbian neural network approach on several high-resolution SAR images. Experimental results show that the proposed approach is quite robust under different imaging conditions and can correctly detect ship targets in a rapid manner.