Traditional CamShift algorithm has been widely applied in the field of video tracking, but it tends to fail in the complex background and occlusion condition. In this paper, we choose color, texture, and the target motion information as features based on traditional CamShift algorithm, and constantly predict the target state of motion combined with the Kalman filter, in order to improve tracking accuracy in the complex background. In the event of occlusion, we use least squares fitting and extrapolation to predict the target location through the priori motion information of the target before occlusion, and re-capture the target after occlusion. The experimental results show that the algorithm can still track the target well in the case of complex background and short-term occlusion and have good real time ability.