Real-time surveillance system detecting persons in complex scenes

Abstract

This paper presents a visual surveillance system that can work in real time under various disturbances, such as changes of lighting and movements of background objects. We propose a robust method of detecting moving objects based on voting in a space-time cube. We have implemented this method on an image processor using multiple DSP boards. Experimental results have shown that the system can detect persons walking with many objects moving in the backgrounds in real time.