Research And Implementation Of The System For Driver Fatigue Detection Based On DSP

Recently, Chinese traffic accidents are in high frequency. The death toll has been ranking first in the world. Driving fatigue is a major factor for traffic accidents. There for, scientific research institutes in many countries have focused on fatigue detection research. The research for improving traffic quality, protecting people's lives and property is of great practical significance.In this paper, we firstly discuss the background and then analyze the main existing fatigue detection algorithms. After analyzing the methods currently used by others, we present eye-locating algorithm based on DSP and dynamic track using alterable rectangular frame and real-time fatigue detection processing. We also present yawning detection algorithm of lip locating and mouth shape analysis based on face geometry character and AAM .The work is described as below.(1) System platform based on DSP is built for driver fatigue detection. After summarizing the state of the art of driver fatigue detection and considering the limitations of existing systems and uptrend for the research, we build the system platform based on DSP.Afterward, the work principle and function implementation for each module are analyzed.(2) The methods of eye locating, dynamic tracking using alterable rectanglar frame and real-time fatigue detection processing on DSP are presented. Eye is located by adopting red-eyed effect generated by adaxial LEDs and preprocessing the image. After that, an alterable rectangular frame is defined to track the eye and extract eye feature in video sequence quickly. Finally, we calculate the value of PERCLOS and count the time in which the value of PERCLOS is greater than the threshold by using sliding time window, and then judge the fatigue status. The experimental results show that our solution has high fatigue detection rate. This solution could be used to monitor the driver fatigue state.(3) Yawning detection algorithm of lip locating and mouth shape analysis based on AAM is presented. We use AAM to build lip shape description model. Because AAM hold description capacity of ample texture information and shape information, it could trace mouth shape change exactly. Then, lip feature points are extracted using lip shape description model. Finally, we calculate the Degree Of Open-moth and count the duration when DOO is greater than threshold to judge whether the driver is fatigue or not. The experimental results show that detection method has high performance. This method could be used for yawning detection to determine driver fatigue state.(4) Driver fatigue detection prototype system based on eye state analysis is designed and implemented on DSP platform. Driver fatigue detection prototype system based on lip locating and mouth shape analysis is designed and implemented by using Visual C++ 6. 0. System testing shows that both two systems could be used to monitor the driver fatigue state.