The Research On Real-Time Detecting System Of The Driver Fatigue

With the stead increase of motor vehicles, more and more attention is given to the road traffic safety, especially to the driver fatigue. Fatigue affects drivers' alertness and ability to drive safely, and many of vehicle collisions result from driver fatigue. Much work has been done in driver fatigue all over the world.Drowsy driver detection methods mainly centralize on the two aspects: based on physiological parameter of driver and based on character of vehicle. However, many methods of the two aspects are too restricted to vehicle or the experience and performance of driver, or having to contact with driver. Of the drowsiness-detection measures and technologies evaluated in this study, PERCLOS is the the only measure of driver fatigue that has been scientifically validated by the National Highway Transportation Administration (NHTSA). The technical definition is"the proportion of time that a subject's eyes were 80% to 100% closed."In this paper , PERCLOS is selected to evaluate driving fatigue , and a drive-fatigue-monitor system based on computer vision technologe is developed. In order to recognize the state of the eyes, the face must be located at first. An algorithm based on two consecutive frames subtraction and back-ground subtraction is chosen to locate the driver's face in gray level images. Then geometry feature information of human face is used to determine the eyes'probable area, in that the eyes'color blocks are detected in YCbCr color-space. According to the geometrical characteristic of the blocks, the closure of the eyes can be calculated, and the calculation of PERCLOS can be resolved. The system detects the driver's alertness level on the basis of PERCLOS, and gives a warning when the driver is in the situation of fatigue.In this paper, a new approach is used to detect the state of human eyes, which is more accurate than the common method. According to the analysis of the experimental data, an multiple threshold approach is proposed to determine the driver fatigue PERCLOS threshold that is used to judge whether the driver fatigue.