The Research And Optimization Of Eye Detection And Tracking Algorithms

With the rapid development and innovation of display technology, in order to avoid using auxiliary equipment while watching 3D images or videos and gain a more comfortable viewing experience, auto-stereoscopic display technologies emerged. Among all these technologies, the grating auto-stereoscopic display technology has become the hottest one and attracts a great quantity of companies and researchers cause its unique advantages. Meanwhile, users would get the images affected seriously by the inevitable crosstalk if they have their head moved while watching because of the fixed and limited best view areas of the grating auto-stereoscopic display technology. In order to solve this problem, an eye-tracking auto-stereoscopic display technology have been proposed, moving the RGB sub-pixels according to the real-time position of the human eyes to reduce the image crosstalk.In this paper, according to the requirements of the eye-tracking grating stereoscopic display system we design and optimize two algorithms on the human eye detection and tracking, and integrate the algorithm into the display system. The content and results are as follows.First of all, make a comprehensive analysis of the current domestic and international eye detection and tracking algorithm, overview the classical algorithm, summarize the advantages and disadvantages and provide a theoretical basis for the research of this paper.Second, overview the AdaBoost training algorithm, the Haar-like feature, the integral image, and the eye detection system based on AdaBoost training algorithm. Meanwhile based on the research of the Kalman filter, an improved detection algorithm is proposed by combining the AdaBoost cascade classifier and the Kalman motion estimation to optimize the effectiveness of the human eye detection and tracking. A series of experiments show that the algorithm improves the detection rate and reduce the detection time.Third, by summarizing the traditional template matching algorithm, an novel template matching algorithm has been proposed according to the requirement of the auto-stereoscopic display system, optimize the selection of the eye template, the calculation of the correlation coefficient, the methods of searching and the obtainment of the thresholds. Then using Matlab to do the experimental simulation and testing the effect of the novel template matching algorithm.Moreover, make a brief slit of the eye-tracking grating auto-stereoscopic display system, combine the display system and the eye detection and tracking system based on AdaBoost cascade classifier and Kalman filter to move the RGB sub-pixels according to the real-time position of the human eyes. Then make the feasibility analysis of the human eye detection system based on the improved template matching algorithm and offer several future research directions.