Automatic Classification of Small Ships in ISAR Images Using 3D Models and Silhouette Matching

Abstract:
An algorithm for automatic classification of Inverse Synthetic Aperture Radar (ISAR) ship images has been implemented and tested on real data. This paper describes the algorithm and presents experimental results. Our approach to automatic classification is based on comparing the silhouette of the target with a training set of silhouettes produced from 3D models. The models are built from photographs using commercial software. The results show that this method has a potential for classification when the height profile of the target is visible in the ISAR images. Test of several image frames simultaneously gives better results than testing the images individually.