SAR ship detection using sea-land segmentation-based convolutional neural network

Abstract

Reliable automatic ship detection in Synthetic Aperture Radar (SAR) imagery plays an important role in the surveillance of maritime activity. Apart from the well-known Spectral Residual (SR) and CFAR detector, there has emerged a novel method for SAR ship detection, based on the deep learning features. Within this paper, we present a framework of Sea-Land Segmentation-based Convolutional Neural Network (SLS-CNN) for ship detection that attempts to combine the SLS-CNN detector, saliency computation and corner features. For this, sea-land segmentation based on the heat map of SR saliency and probability distribution of the corner is applied, which is followed by SLS-CNN detector, and a final merged minimum bounding rectangles. The framework has been tested and assessed on ALOS PALSAR and TerraSAR-X imagery. Experimental results on representative SAR images of different kinds of ships demonstrate the efficiency and robustness of our proposed SLS-CNN detector.