A Multi-object Motion-tracking Method for Video Surveillance

In this paper, a multi-object motion-tracking method based both on region and feature tracking is proposed for the purpose of real-time tracking in video surveillance system. Moving object is detected through foreground detection. Then we extract three features of each moving objects such as centroid, area, and average luminance. At last, the similarity function is applied to tracking. It is proved that the method has good performance under dynamic circumstances for real-time tracking.