Abstract: The man-machine interaction approach based on line-of-sight tracking is characterized by promptness and celerity, and it can be used to achieve the interaction between human sight and computer. This paper builds one kind of helmet line-of-sight tracking system. A cascade pupil localization method is proposed in order to eliminate various obstruction factors. First, the pupil area is retrieved using its binary image and the minimum outer ellipse is calculated. Second, the overlapping portion of the pupil area and the minimum outer ellipse is retrieved and used to calculate the ratio of the overlapping area and pupil area. If the ratio is under the threshold value, the pupil is uncovered; otherwise it is covered. The particle swarm optimization (PSO) and ellipse difference are used to search the best pupil location. Experimental results show that our method is a fast and robust method, and an effective strategy for pupil localization in complex environments.

Key words: line-of-sight tracking, pupil localization, covered pupil, particle swarm optimization, ellipse difference