Research Of Multi-video Based Driver Fatigue Surveillant System

With the development of highway and the increasing speed,traffic accidents have become a serious problem for the society which threatens people's life and safety. According to the statistics from the Department of Transportation, fatigue driving ranks the third in the factors causing traffic accidents, and the first in fetal traffic accidents. Thus, developing a system for monitoring the driver's level of vigilance and alerting the driver when he fatigues is essential to avoid traffic accidents caused by fatigue driving.Now the video based driver fatigue surveillant system is popular because of its real-time capability and veracity.Based on the analysis of current research status in domestic and abroad, this dissertation studies key technologies of the video based driver fatigue surveillant system. The core content of this paper is that realizing and improving algorithms such as face location,eye location,eye information extraction,lips information extraction and fatigue level calculation. The main contributions of this research are:The multi-video based photo capture equipment is designed, the efficient algorithm AdaBoost is adopted to detect and locate faces. We improve the complexion modeling method; combine the face detecting algorithms based on complexion and AdaBoost, then present a face re-locating algorithm in the course of surveillance. The AdaBoost is still used in the key point eye location, however to improve the popular tracking algorithm mean shift, we propose an adaptive kernel-bindwidth updating method and a new adaptive target occlusion estimating method based on face isoceles triangle, which makes the continuous tracking come true. At last, after eyes and the mouth are located successefully, the PERCLOS considered as the current most valid fatigue level index and the yawn are selected to evaluate the driver's fatigue level.A lot of experiments are designed and carried out due to the variety of actual applications which including different lighting conditions, different people, and different face poses, etc. The experimental results show that, the proposed system can judge the driver's fatigue status precisely.