A Hough based eye tracking algorithm without on-site calibration

摘要

This paper describes an algorithm for stable eye movement detection without on-site calibration even when the relative position of head and camera changes. In the algorithm, after detecting the 3-dimensional coordinates of cornea curvature center, Rough transforms is used, and pupil shape is detected with high precision. Then, the rotation angle of pupil is calculated based on the eyeball model, and the eye movement direction is detected from the rotation angle. Although the individual eyeball informations (the cornea curvature radius and the distance from the cornea curvature center to the pupil) are used, the on-site calibration is unnecessary once they are measured. The experimental results show the proposed method is effective: the eye direction can be measured in the accuracy of almost same level as the involuntary eye movement, without on-site calibration even when camera position moves.