Research And Implementation Of Fatigue Detection Technology Based On Eye Feature

With the rapid development of the world economy, the car has become an indispensable transportation in social life and has made great contribution for social economy. However, with the development of transportation industry, traffic accidents have become a serious problem in the current countries, fatigue driving is a main factor in traffic accidents. So it was practical to research and implementation effective algorithms of fatigue detection.On the basis of researching the current situation of fatigue driving and reference the existing literature both at home and abroad, this thesis systemically put forward concerning fatigue detection’s overall design scheme. Firstly, in the detection stage, face detection was implemented based on the face Adaboost cascade classifier, and the driver’s face was found after judging all the faces,then eye detection was implemented based on the eye Adaboost cascade classifier in driver’s face region. Secondly, in the tracking stage, eye regions were predicted through the displacement of the three frames before, and then template matching was used in these regions. However, because the traditional template matching algorithm’s robust is not good, the accumulated error continuously and the eye blinking frequently led to the eye tracking lost, facing with its problems, updating template was introduced in this thesis. Finally, in the fatigue judging stage, on the basis of accurate tracking, the height and width of each eye were obtained, the TOOE was calculated, and the eye feature was judged, then driver’s fatigue was judged by calculation two fatigue parameters (PERCLOS and blink frequency). This method improved the stability of the driver’s fatigue detection caused by the single fatigue parameter, and ensured the accuracy of the fatigue detection.On PC, this thesis used VS2008development environment and C#programming language, and then implemented this algorithm was based on Open CV. Test has been done on real-time and accuracy under different prediction method, light, movement speed, rotation angles and fringe interference. The test result indicates that in the normal sunlight and the pixel-level video, the driver’s detection algorithms introduced by this thesis can be precisely implemented on time in every process, and can accurately reflect the driver’s fatigue.