A surveillance method for driver's fatigue and distraction based on machine vision

Abstract

Driver fatigue and distraction monitoring and warning system play an important role in reducing traffic accident and improving road safety. Using machine vision to monitor driver's driving state has become a research focus in safety driving assist system. A new algorithm is proposed which includes driver's face location based on gray variance, and segmentation of separate areas of facial organs using the projection curve pole position. The eye contour state can be got from OSTU calculation and contour extraction in each separate region. Calculation method of driver's facial rotation angle is presented based on the analysis of driver's head rotation. Different driver attention monitoring methods were used according to different head rotation angles. PERCLOS is used to monitor fatigue state when the driver face forward, while facial orientation is used to monitor distraction state when driver face deflection angle. Results show that the processing algorithm cost 32ms in average, so it has good real-time performance and high accuracy rate.