Research On Fatigue Detection Using Eye Information

Fatigue driving is gradually becoming one of the main causes of traffic accidents, which is called“the principal murder”. So fatigue detection can prevent and reduce traffic accidents, which is good for the demos’security and society’s economy.The thesis simulates the driving environment to detect fatigue based on eye information. The thesis uses MATLAB to finish the process of face detection, face tracking, eyes location and state’s distinguishing and fatigue judgment. The main work is as follows:(1) After face location using the face skin color model combining with the characteristic of the human face, the thesis deals with the face tracking in succession, based on an algorithm for real-time face tracking which combined Mean Shift algorithm and Kalman filter. Of this part, we added template updating to the traditional Mean Shift algorithm in order to adapt the change of face’s distance and angles, and the Kalman filter’s residual equation is used to solve the problem of face’s shade. The experiment revealed the method can adapt to the change of face and its accurate probability can up to 86.8%.(2)The thesis picked up a method based on twice horizontal integral projection to locate eye and distinguish its states. The method used horizontal integral projection to locate eye roughly based on the face’s extraction, then get the width of open and closed eyes also by horizontal integral projection. At last the PERCLOS method is used to judge the fatigue according to the eye width’change of video.(3) We use fuzzy logic method to simulate the fatigue judgment based on blink’s time and frequency in MATLAB’s fuzzy toolbox. The result reveals the reliability probability is relatively higher. Finally a kind of fatigue detection’s system model based on new wireless network ZigBee technology is proposed to save the wiring’s space of car and contact with the outside by wireless.