Traditional mean shift will fail when the goal’s color is similar to the background or the goal is blocked.An improved mean shift tracking algorithm is presented.RGB color space is coverted into HSI color space.The interference is overcomed when the goal’s color is similar to the background.When the goal is blocked

a time-invariant system is used to describe the movement of the target during a short time sequences.Through Kalman filter

this system is identified so as to make it have the ability of estimating the coming states.Experimental results demonstrate that the proposed approach can accurately track the goal in complicated scenes

and can deal with 39 pictures per second with a Intel 1.73 GHz CPU.