The Research And Implementation Of Real-time Eye-tracking System In Desktop

The paper first introduces the research and application of eye tracking technology and its background as well as its necessity,reviews the development process of the technology at home and abroad,then summarizes the main eye tracking technology and its principles,and at last compares the advantages and disadvantages both these technologies.Accurate positioning and tracking of the pupil is one critical step in the eye tracking system that it can work stably.In this paper,a strong robust real-time detection and tracking algorithm of pupil is proposed,especially in the case of wearing glasses.The algorithm is based on bright and dark pupil image,it is firstly by collecting the bright and dark pupil image,and then these two images are inverse binarized.After that,we can filter the area that can't be the area of pupil in the black inverse binarized picture.Then the imgages of dark and bright of inverse binarized are differencing.Finally,the pupil is roughly positioned in the differential image.The region of interest are then selected in the dark pupil according to the previous positioned pupil center,and elliptic fitting is performed in the region of interest around.Afterwards,the center of the pupil and the center of the spot are calculated.In addition,the result of image binarization directly affects the last feature extraction,but the traditional binaration method would be serious affected by its two value threshold.So that the scope of application are limited.Therefore,this paper first based on the function of adaptive segmentation method in OPENCV,then counting the relationship between the mean gray value and the last parameter of adaptive-threshold function.Finally,the last parameter of the adaptive binarization function changes with the average gray around the eye.Through a large number of experiments show that the algorithm not only in the ideal situations to run stablely,but in complex situations such as wearing glasses and changing the light or varing the gesture of face,the algorithm can also perform well.Finally,the system implementation and performance test are carried out in PC.The system consists of two parts: hardware platform and software.The hardware platform mainly composed of USB3.0 dual image acquisition card,camera,infrared light and PC.The software part is divided into acquisition module and application module,which mainly completes the extraction of image feature points and the placement of line of sight.Through the experiment,the error of the single-camera single-light system in the laboratory is about0.8 degrees.