A realtime tracking system for the fast moving object on the complex backgroun d is proposed. The Markov random filed (MRF) model based background subtraction algorithm is used to detect the changing pixels and track the moving object. The prior probability of the segmentation mask is modeled by using MRF

and the object tracking task is translated into the maximum aposterior (MA P) problem. Experimental results show that the method is efficient at both offli ne and online moving objects on simple and complex background.