Research on Algorithm of Eye Tracking in Identification of Human Eye

摘要

In video stream, a method was mentioned that combining gray level projection and human iris saturation information based on HIS color space with template matching and Kalman filter in order to improve the robustness of human eye positioning and tracking, accuracy and detection speed, and improve the applicability of the algorithm. First, the gray level projection method to locate eyebrow regions was adopted, then human iris saturation information was used to locate the eyes accurately. Next, according to the characteristics of Kalman filter, the location of the human eye to appear in the next frame would be estimated, and template matching would run in limit range in order to determine the exact location of the eye. The experimental results show that the right accuracy of eye positioning is 85%, positioning time is 0.0726s, the rate of eye tracking could reach 17 f/s, the accuracy of eye tracking is 87.5%.