Research And Implementation Of Drivers’ Fatigue Detection Based On Computer Vision

With the rapid development of the world economy, the processing capacity of cars has been daily on the increase. Traffic accidents caused by drivers’ fatigue driving become more and more. In order to guarantee driving safety and guard against occurrence of traffic accidents, it is practical to study a method used to effectively detect drivers’fatigue and give warning timely.This thesis systemically studied and analyzed the present situation of the fatigue driving both at home and abroad, in special, it studied and implemented the driver fatigue detection algorithm based on computer vision. This method mainly consists of face detection, eye location and track, collecting of eye features and judgment of fatigue, and others. In the process of eye tracking, the internal and external corners of the eye were obtained through Harris corner detection algorithm, and also KLT algorithm was adopted. However, in the actual process of eye tracking, hypothesis about the consistency of gray distribution between feature windows and small displacement between adherent frames of KLT algorithm can not be satisfied, this gave a result that the eye tracking appeared a little displacement with the feature of accumulative effect. Facing with such problem, the KLT algorithm was improved through adjusting the tracking result according to the obtained eye corners in the process of extracting the features of eyes in order to eliminate the tracking deviation. Otherwise, in the process of extracting the features of eyes, in order to guarantee the extract of eyes features more simple and objective, this thesis gave out three extracting method based on three situations of eyes. Firstly, this method used Canny algorithm to get the openness of the eye, then to make sure the situation of eyes based on the openness of the eye, at last to calculate the features of eyes according to successive situations of eyes. When image acquisition was failed or eye features can not be extracted through images, it dynamically makes sure how to deal with the frame according to drivers’fatigue degree.This thesis used VS2008development environment on PC and implemented this algorithm based on Open CV. Test has been done on speed and robustness under different conditions of illumination, noisy density, and head gestures. The result indicates that on the usual sunlight and noise, the fatigue testing method introduced by this thesis can precisely reflect different situations of the drivers in every process.