A real-time fatigue driving detection system design and implementation

A reliable and real-time fatigue driving detection system is very important for traffic safety. Till now, various methods have been presented to improve the accuracy and robustness, however, few papers showed the efficiency of the system or the feasibility of real-time processing in embedded system. Considering the efficiency and reliability, a real-time fatigue driving detection system is presented in this paper. Firstly, Haar-like features and AdaBoosted classifiers are adopted for real-time face detection. The selection of the optimal sampling ratio and the minimum possible object size is discussed for high detection accuracy and computational efficiency. Secondly, eye location and iris positioning are applied to detected faces. We propose to use the area proportion that the iris takes in its minimum circumscribed circle as the indicator for judging the state of the eye, which is more convenient and has good robustness to distance. Finally, the percentage of eye closure (PERCLOS) is used as the criteria to determine whether the driver is tired. The embedded platform implementation of the proposed system turns out to be efficient.