G. Sun, Y. Jin, Z. Li, F. Zhang, L. Jia

Pages: 51-64

Abstract
Fatigue driving is a major cause of traffic accidents. Real-time driving fatigue detection can be achieved by monitoring the variation of driver’s driving behavior, such as nodding phenomenon(it can be detect by observing head position changes) can be applied to driver drowsiness detection. In the traditional algorithm based on head statement detection, however, hardware sensors (such as HPS) are indispensible to feedback head position information. The traditional algorithms using hardware sensors are difficult to gain popularity given its high installation and maintenance costs. To improve this, with image processing technology, lots of vision-based algorithms are proposed to determine head position for driving fatigue detection. However, they all need to detect face organ to determine the head status, which increase the calculation cost of the system. In this paper, face organ detection is replaced by hair detection and head status recognition is achieved based on the variation of position relation between face and hair under different posture. The algorithm presented in this paper, since there is no need to detect face organ, can be applicable to low-resolution images and blur images.

Keywords: driving fatigue detection; head position; face detection; hair detection