Research And Optimized Implementation Of Eye Tracking Technology

Eye tracking, also called line of sight tracking systems is one of the keytechnologies of intelligent human-computer interface in the future. It has a broaddevelopment prospects in both military and civilian areas, so it becomes front directionwith great attention in resent years.At first, this thesis introduces the background of eye tracking system, and provesthe necessity of using this technology in high-tech industries research.Next this thesis proposes systematic framework, introduces the calibration methodused, and describes improving methods of pupil center localization based on start burstspecifically: preprocessing collected gray eyes picture at first, then calculating theapproximate location of the pupil center according to adaptive threshold binary imageobtained by histogram, and looking for the corneal reflection spots central positionaccording to the location of the pupil center. At the same time, using extending radialsmethod detects pupil figure point, removing false figure points generated by eyelids,eyelashes and reflection spots by passing window to figure points. Then selecting sixpoints randomly from different areas to fit ellipse and set double threshold value toobtain best fitting Ellipse. This thesis presents a fast method to determine blinking,using the best optimum linear approach to deal with blinking, which is simple and canmeet the real-time requirements. Analyzing the idea accuracy of head-mounted systemis 0.1°. The accuracy of testing pupil center by experiment is about 0.18°.Finally this thesis discusses the problems needed attention when porting eyetracking approach to CCS. Using code tuning, library functions extension, compilingand other methods to optimize the processing time from 8s to 80ms, which increases thespeed about 100 times .