Design On Eye Location And Tracking System

Computer vision is a challenging field,its huge development potential has attractedmany researchers explore and research. With the continuous improvement of computerand the popularity of electronic products, more and more researchers have devoted to animportant research direction which is the human eye detection in the field of computervision. Eye tracking can be used to achieve fatigue test, the disabled interaction, variousapplications and so on.In this paper, the main research contents can be divided into three parts such asface detection, eye detection and eye tracking. On the basis of in the related algorithm,the tracking method is suitable for real-time face detection and eye detection.Software ismainly VC++6.0with OpenCV,which set up a human eye tracking system by callingthe OpenCV library function and greatly shorten the development cycle of the system.Firstly, in this paper, which introduce the Boosting algorithm and AdaBoostalgorithm in-depthly. This paper adopted the continuous Adaboost classifier for thehuman eye detection. First of all,,this paper detect and locate the human eye to the inputimage on the basis of face detection. The pupil location as the human eye tracking isvery important step. The black pupil area in the human eye is relatively stable imageinformation.So,this article finally locates eye pupil, lays the foundation for the humaneye tracking. By continuous Adaboost algorithm based on dynamic image trackingexperiment, which is difficult to meet the real-time requirements of the system. So inorder to improve the speed of real-time detection, the article introduces the CamShifttracking algorithm.CamShift algorithm must select the tracking target before trackingeverytime,which is a kind of semi-automatic tracking algorithm.But after AdaBoostalgorithm the idea was combined with CamShift algorithm, the AdaBoost algorithmdetected the eye area and the pupil position as initialized search window, which cansolve the problem of the semi-automatic tracking.The Meanshift algorithm correct theShortcomings, adaptive adjustment of tracking window size and the target improve thetracking performance. But in the process of Camshift tracking, the global image ofreverse projection can increase the computational complexity, therefore,the originalCamshift algorithm to reverse the global image projection to local reverse projection ofthe target, which reduce the computational tracking process complexity and result in amore accurate tracking results.Usingthe pupil center of human eye and center of offset between two eyes, eye gaze is derived, thus realize the eye tracking and control mouse.In the process of the human eye trackin, the improved CamShift algorithm whichcan conduct the human eye tracking, namely, after the first frame video image capture,the cascade classifier based on AdaBoost algorithm of input image of face detection andeye detection and to locate the pupil center.which is as a benchmark to determine initialsearch window, conveniently solute the problem of the semi-automatic tracking. Andafter the detection of the human eye, the eye area is extracted through the study of thebinarization processing of the human eye and eye image on integral map projection andjudg the state of the human eye Simply.In this paper, the development of the human eye tracking and human-computerinteraction system realize the real-time tracking human eye detection, and use thehuman eye for human-computer interaction. This system does not have a high demandfor hardware and has very good robustness to certain illumination changes, large sidefaces and pitch change.