Estimation of Driver Drowsiness Using Blood Perfusion Analysis of Facial Thermal Images in a Driving Simulator

Background and Objective: Driver drowsiness is a major cause of fatal driving accidents worldwide, which can be prevented by an early detection. Driver drowsiness results in reduction of blood flow and change in facial thermal patterns. In this paper, a drowsiness detection system was designed to estimate the level of drowsiness by analyzing facial skin temperature patterns. #xD; Materials and Methods: 12 subjects drove on a highway in a driving simulator, while their facial temperature image was captured by a thermal camera. The subjects #x2019; drowsiness levels were independently estimated by the observer rating of drowsiness (ORD) method. Facial blood vessels were located by a fourstep algorithm and tracked in each frame. The mean value of the pixel intensities of the facial blood regions formed a sequence of data. Variations of facial skin temperature signals were investigated for three main blood vessels. #xD; Results: Facial skin temperature decreased at three levels of sleepiness from wakefulness to extreme drowsiness. Temperature near the supratrochlear, angular, and facial arteries decreased by 0.54 #xBA;C, 0.33 #xBA;C, and 0.32 #xBA;C from wakefulness to extreme drowsiness, respectively. All subjects except one experienced skin temperature decrease from wakefulness to extreme drowsiness. #xD; Conclusion: It was shown that the drivers #x2019; facial blood vessels temperature decreased as the level of drowsiness increased. Driver drowsiness can be detected by monitoring the change in the facial skin temperature signal during driving. Facial thermal imaging is a promising nonintrusive method to detect driver drowsiness.