Classification of Ships in TerraSAR-X Images Based on 3D Models and Silhouette Matching

This paper demonstrates the feasibility of target recognition of ships in harbour from high-resolution satellite Synthetic Aperture Radar (SAR) images. The silhouette is found to have a potential as a feature for target recognition due to the fact that layover makes the height profile of the ships visible in the images. In the approach demonstrated in this paper, the silhouettes extracted from a SAR image are compared with silhouettes of 3D models which can be built from photographs. The experimental results with data from TerraSAR-X show discrimination between ships with comparable size and shape.