Research Of Eyeball Location Tracking System Based On Machine Vision

With the rapid development of era,more and more electronic products appear in people’s daily life,the extent of people using eyes is increasing day by day,with that comes the frequency of eye diseases,so that being timely and effectively diagnosing eye diseases is particularly important.In the process of examining pathological changes of eyes,the microscopic photos of retinal cells,which are analysed to estimate whether the eye lesions occur,are obtained with the method of the eye film.Due to the low efficiency and poor quality of artificial filming,an automatic positioning tracking device of automatically looking for eyeball position is obvious.According to the corneal endothelial cell analyzer manually locating and tracking eye,this paper put forward a kind of machine vision eye locating tracking system based on embedded platform,achieved automatic positioning tracking of eye with the method of image processing combined with infrared distance measurement.The research direction was determined based on the research of the development situation at home and abroad,the feasibility of research was illustrated,the basic technical route was determined through the analysis of the functional requirements of system.The development environment was set up by the way of establishing an embedded Linux 3.5 kernel through Exynos4412.An improved target tracking algorithm was raised through the study of the existing traditional positioning tracking algorithm,the target was detected and the target edge was extracted through the continuous frame-difference method,four cross positioning method is used to obtain two-dimensional coordinate of the eyeball,combining with the depth information collected by the infrared distance sensor,the three-dimensional coordinate of the eyeball was determined.And according to the positioning information,the stepping motor control parameters were updated timely,with which the camera was controlled to track the eyeball movement trajectory,as a result,the locating and tracking of eyeball were achieved.The algorithm was realized using the C language programming,the speed of the algorithm was greatly improved under the premise of the precision,reliability and robustness of the algorithm being not affected,which made it more suitable for embedded platform.The transplantation of algorithm from the PC platform to the embedded platform was accomplished,and the precision,real-time performance and reliability of the system were tested.The experiment proves that the positioning error of x/y axis is not more than 4 pixels in view of the eyeball and other objects of different shapes,the positioning error of z axis is not more than 2mm,the localization algorithm takes less than 2 ms,and the tracking trajectory is largely in line with the target motion trajectory,the tracking takes no more than 7s under the premise of the distance between the target and the camera being not more than 160 mm.After demonstration,it is obvious that the system has high precision,reliability and real-time performance,and also has wide application areas and broad market prospects.