Real Time Human Eye Detection Algorithm And Embedded Systems Research

Eye is one of the most prominent features of human face. In artificial intelligence, human-computer interaction, pattern recognition and other fields of study, the method of eye detection is a crucial issue. We can obtain people’s psychological activity, demand, cognitive processes and emotions etc., through the eyes of detection. Human eye detection theory is mature, which has been used in the practical phase of human life. How we can better apply to the human eye detection to the embedded platform is still a difficult problem.In this paper, we present a new simplified eye model, based on the "Gullstrand’s principle eye". We analyze several key factors to generate bright pupil using the eye model, and put forward the method of bright pupil acquisition in the actual eye detection applications. While we design an eye bright pupil image acquisition system. Based on production conditions of bright pupil, we start from the infrared light source brightness, delay, equilibrium and so on. Through the system test and the analysis of calculation results, we effectively improve the quality of human eye image to be detected.Through analyzing the principle of embedded platform features and AdaBoost algorithm, we optimize the AdaBoost algorithm:the speed of the algorithm is increased by 218%, while accurate rate reach to 80%, using region selection algorithm optimization of AdaBoost face detection process. In region selection optimization, skin color region selection algorithm is used to segment skin regions of the color image, the OTSU algorithm is used to segment future (human body) of gray level image. In embedded platform adaptive optimization of AdaBoost algorithm, haar characteristics are optimized respectively in the calculation of the floating point operations and square root operation, The speed of the AdaBoost algorithm is improved of 81% and 32% respectively, the accuracy rate is basically unchanged. The research work in this paper can effectively improve the face detection speed and accuracy in the embedded platform.Combinating of the above research work, we put forward a real-time eye detection system in embedded platform. In this paper, we analyze the problems in the embedded platform of eye detection algorithm, and give the eye detection scheme as follows:We roughly locate the human eyes using projection operation. The binary image of the eye area is used noise reduction algorithms to remove smaller bright spots, which could distinguish bright pupil and noise. Then We use region growing algorithm to posite the bright pupil area. Finally, we use CamShift algorithm to track human eye. In summary, we achieve a real-time detection system of the human eye in embedded platform...