Differential trajectory tracking with automatic learning of background reconstruction

Nowadays, trajectory tracking technology is widely used in many outdoor applications, such as intelligent traffic and video surveillance. However, most of trajectory-tracking technologies rely on a static background, which is hard to obtain in many situations. Obviously, these methods are out of action in the case of dynamic background. In this paper, a novel trajectory tracking method is presented, which is implemented with a new background reconstruction algorithm. Firstly, the background is assumed to be a blank scene. Then, the background is reconstructed by means of video detection that places moving objects in the scene. Finally, real-time trajectories of moving objects are computed based on the reconstructed background. Experimental results show its robustness and practicability even in a cluttered background.