The Research And Implemention Of Desktop Eye Tracking System

Eye tracking technology has been developed rapidly in recent years,and has been widely used in psychology research and commercial sector.More and more researchers in different fields participate in this project.in the future it will become one of the most important way of human-computer interaction.First of all, this paper introduces the research background of eye tracking technology and the development course, summarizes the domestic and foreign studies of eye movement tracking method.Then,this paper implements a desktop gaze tracking system on a PC.Using Matlab to do the camera calibration makes the whole calibration process more convenient.In the eye procession, this paper USES the bright-dark pupil technology, which makes the extraction of pupil more accurate.By using OpenCV,It takes about 5ms to process a frame of image,which satisfy the requirement of the whole system real-time performance considering the influence of various error,this paper adopte four point correction algorithm, which makes the average accuracy of the whole system about 0.7°.For using 3d model, the system can allow users to move within a certain range of freedom head.Last,This paper introduces multiple results problems in which to use the Newton iterative method to solve the model.The paper analyse the number of model solution and the geometric relationships between these solutions.Through that,we can find the correct solution of the constraints and determine a reasonable initial iteration value.So the whole System can not only finish the solution of the model at about lms,but also reach the right solution in the case of a free head moving. This paper also analyzes the error of the system,the results show that the effection of the eye parameter error effectthe precision of the system is not big, but the system of real-time input error influence is very big. By setting the three-dimensional position of the light source the impact of input errors can be greatly reduced.