The detection of naval vessels by fusion of edge and color background models

Abstract

The detection of naval vessels in open water is a difficult challenge, with many applications from harbor monitoring to proximity security systems for ocean-faring ships. Background modelling is an effective method of detecting candidate targets, and we describe a framework that improves the operational robustness by the interaction of two different Mixture of Gaussian (MoG) models. These results are then processed by a tracking system that accurately detects moving vessels. The algorithm performs well in varying environmental conditions with good real-time performance characteristics.