Abstract: In order to realize the requirement of precise eye tracking in clinical, a pupil center location algorithm based on the least square method is proposed. First, the eye image was captured by the camera under the infrared light, and then the two-valued image was obtained after preprocessing. Use the number of pixels that form the outline which was Extracting from the two-valued image to carry out the rough filtration of the pupil. The ellipse curve was fitting by the randomly select 6 pixels on the filtered contours, and then calculated the variance of distance between the center and the edge of the ellipse. The ellipse center is ellipse center, where the variance of the distance is Minimum. The algorithm has advantage of accurate identification the pupil center when there is white spot interference and squint. Experiments show that the pupil center can be located quickly and real-timely by this algorithm.