Structure Design Of The Head-mounted Eye Tracking System And The Acquiring Method Of Eye Information

Eye-tracking technology is a key technology of intelligent eye-movement operating system,and it is one of the key technologies of intelligent human-machine interface.At the early stage of development of eye-tracking technology,the measurement method is poor and has great interference to users.In recent years,with the development of machine vision and image processing technology,the optical recording method based on computer vision has become the hot spot of current research.There are three common computer-based optical recording methods: single camera without light source system,single camera with multi-light system and multi-camera with multi-light source system.The system without light source mainly relies on image processing algorithms to detect eye characteristics,its structure is simple,but the algorithm is complex.Single camera with multi-light system and multi-camera with multi-light source system generally based on pupil-corneal reflection method to track the human eye,while theirs structure and calibration process are complex.Aiming at the above problems,this paper builds a head-mounted small eye-tracking system based on binocular stereoscopic vision.The head-mounted small eye-tracking system is simple,easy-using,low-cost and accurate.The head-mounted small eye-tracking system consists a near-infrared light source,a binocular camera,a scene camera and a reflector.With the help of reflection mirror,binocular camera capture human eye images,while the scene camera captures image of scene.The binocular and scene cameras are calibrated to unify the human eye image and the scene image into the same coordinate system.As the scene camera and binocular camera position special,a method utilizing 2-axis turntable for multi-camera calibration without public FOV(field of view)has been proposed in this paper.Eye-tracking is the core and key of the whole line-of-sight tracking system.The paper is based on the definition “the normal direction of the circle plane of the pupil edge in single eye is the optical axis” to exact eye information.First,the pupil sub-pixel edge detection and the pupil center localization are performed in the infrared image of the eyes;then,the corner area is roughly determined according to the pupil position and its accurate position can be extracted according to the gray value accumulated features;finally,according to the characteristics of the pupil and the corner of the eye to determine the same name points of two pictures,and then with the camera calibration results,the line of sight direction can be solved.Human eye images and the line of sight direction can be accurately captured by the head-mounted small tracking system based on binocular stereo vision described in this paper.It solves the problem that the existing eye tracking system has complicate structure and cumbersome calibration process.