Through analyzing the advantages and disadvantages of the template matching method, Kalman Filter and Mean Shift algorithm, we proposed the method of combining them. We use the template matching method to initialize the irises positions, extract the eyes area, and then combined Kalman filter with Mean Shift algorithm to position and track irises accurately. The method also added to the average speed of the targets, it not only could be used as the standard if the irises are affected by the background, but also could be used as the irises speed when eyes are blocked, and then forecast the goals possible area in the next frame. This method is able to overcome the problems about irises tracking failure when there are tilt angle, deflection angle and pitch angle of the head in a certain range, and the object occlusion and background interference problems, which compressed the amount of computation and improved robustness of the iris tracking.