Gaze Tracking Algorithm Based on Projective Mapping Correction and Gaze Point Compensation in Natural Light

Abstract

Gaze tracking in natural light is a challenging task due to restricted head movement and low accuracy. This paper presents a novel gaze tracking algorithm based on projective mapping correction and fixation point compensation to address this problem. Proposed algorithm utilizes the vector between iris center and eye corners as binocular tracking features to estimation the point of gaze, however, the features of iris center and eye corner are very vulnerable to head movements. We analyze the positional relationship of eye features on image while head moving, and propose a method based on projective mapping correction to adjust the feature vector to reduce the impact of head movement. Then the estimation result is compensated by a support vector regression model which created on the relationship between head movement and point of gaze deviation. The gaze mapping model which can handle unconstrained head movements is constructed. The experiments show that the gaze tracking algorithm proposed can effectively reduce the impact of head movement, and improve the fixation point estimation precision in natural light.