Moving object detection and tracking in video surveillance system

Abstract

This paper presents solutions to some problems in real-time video surveillance system such as robust object detection in illumination mutation situation and merge/split event in multi-target tracking. We propose a new moving object detection approach. Our approach can handle the problems result from illumination sudden changes in real scene by combining Gaussian Mixture Model algorithm with temporal frame difference. For multi-target tracking, we use Mean Shift algorithm and introduce a "record" strategy to handle targets merge and split. Experimental results demonstrate the effectiveness of our proposed approach.