| The main cause of traffic accident—drowsy driving, can be avoided by surveying the fatigue level of the drivers. The"Video based human fatigue level surveillant system"can reduce the traffic harms that caused by the drowsy driving evidently.The following topics are discussed in this dissertation: key techniques are researched and system architecture is designed. The system analyses the biological fatigue signals of human beings, collect the real-time driver's presentational status, and process the signals and status automatically by the proposed algorithm. After this process, the integrated bright pupil effect based driver fatigue level surveillant system has been achieved. The main contributions of this research are:Specialized video capture equipment is designed. The bright pupil effect is realized by append a special infrared light to the system. A SVM based eye-detection method is proposed: the first step, human face region is extracted in the YCgCr color space in order to reduce the data processing, the second step, template convolve and threshold methods are used to enhance the eye features and remove non-eye regions. At the third step, the extracted candidate eye objects are processed by the trained SVM, the real eyes region are extracted and positioned precisely. A Kalman filter and Meanshift trace fused eyes-tracing framework is proposed, under this framework, the drivers'eyes are precisely traced, it proves the tracing algorithm is correct and robust. The drivers fatigue level evaluation index is proposed based on the PERCLOS, AECS indices. Also, the index self detection method and drivers'fatigue status integrated evaluation model are established.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, wearing a glass or not, etc. the experimental results show that, the proposed system can judge the drivers fatigue status precisely. |
PDF Full Text Request