Research On Eye Tracking Technique Based On Pupil Corneal Reflection Method

Eye tracking technology makes use of the movement informationof human eyes to control equipment and solved the equipment operation problem of hands disabled person or hands occupied operators. Nowadays, eye tracking received widely attentions of researchers both in china and abroad. As one of intelligent human-computer interaction technologies, the development potential of eye tracking is very considerable and the application prospect of eye tracking in many fields such as military technology and intelligent machines is very bright.In terms of current eye tracking systems, despite its technology advantage, the tracking accuracy and user freedom still cannot satisfy the actual application requirement because of the physiological complexity of human eyes and demands of head moving. This thesis expounds the theory of eye tracking and researched related algorithms based on the research platform of eye tracking system which applied pupil corneal reflection method. To further improve the estimation accuracy and head moving freedom, this thesis mainly merges some new ideas and focuses on the algorithms in gaze point estimation. The main research in this thesis is shown as following :First, aiming at the problems in traditional gaze point estimation algorithms, this thesis gives an angle mapping based gaze point estimation schemewhich utilizes human eyes characteristics and visual imaging theory. In this algorithm the gaze point is estimated with the angle mapping relationships between the position of pupil center relative to corneal reflection points and the positions of gaze point relative to infrared light sources.Secondly, on the foundation of the gaze estimation algorithm based on angle mapping, this thesis analyzes the impact on estimation results of head moving in detail and afterward verifies the max estimated error of different head moving directions in the allowed range of this algorithm.At last, to deal with the influence of complicated physiological structure of human eyeball on gaze point estimation results, a gaze point estimation error analysis model is built and a gaze point estimation error compensation solution based on arc of eyeball is thought out. Meanwhile, a nonlinear mapping algorithm is adopted to compensate the estimation error caused by eye optical axis error and improves system tracking precision.The simulation shows that the system presented in this thesis has high precision and degree of head moving freedom and the average estimated error in horizontal and vertical direction is less than 1cm. This eye tracking system can be well balance the relationship of head moving freedom, eye tracking precision and system cost.