A Self-organizing Neural System for Background and Foreground Modeling

In this paper we propose a system that is able to detect moving objects in digital image sequences taken from stationary cameras and to distinguish wether they have eventually stopped in the scene. Our approach is based on self organization through artificial neural networks to construct a model of the scene background that can handle scenes containing moving backgrounds or gradual illumination variations, and models of stopped foreground layers that help in distinguishing between moving and stopped foreground regions, leading to an initial segmentation of scene objects. Experimental results are presented for color video sequences that represent typical situations critical for video surveillance sysytems.