To extract the moving target in video sequence fast and exactly,a self-adaptive background estimation technique was developed.Because of the different time-variability between background and foreground,a stability matrix was constructed,and according to its changes the background points could be distinguished from the foreground points.By setting two saturation values to the stability matrix,the algorithm could perceive the saltation of background automatically and update the background image quickly.Besides,in order to register the dynamic image sequences some local features were selected as templates,and projection matching technique was used to simplify the calculation and quicken the offset estimation.The experiment results show this algorithm has a high rate of convergence that only 10 frame are needed to get a stable background image,and the whole time is only 35 ms.So it meet the need of a processing ability of 25 frame/s in engineering.