Mean-Shift tracking algorithm is widely used in vision tracking field because its low computarion complexity, better tracking effect and good real-time performance. Mean-Shift tracking principle is analyzed particularly. The moving object tracking course and realization mthod based on Mean-Shift are proposed. At last, experiments are taken using human sequence images in complex background. The experiment results indicate that this method takes on the better tracking performance, and better robust for transitory small occulusion.