Research On Detection Technique Of Motor Driver Fatigue

At present ,more traffic accidents which were caused of driver fatigue have taken place and how to avoid driver fatigue has become increasingly a cause for concern. In this paper, the software of fatigue detection is investigated. The software takes advantage of image processing technology and PERCLOS(percentage of eyelid closure over the pupil over time) method to detect motor driver's eyes and then comput the fatigue state. The software will alarm driver when drive fatigue is detected and accidents will be avoided in effect. Four aspects are included: eye locating,eye's state judgement,eye tracking and PERCLOS is used to determine fatigue state. In this paper ,8-bitmaps are used, following are the methods of four above aspects:(1) Eye location: Two steps are proposed to locate eye, one is face locating and the other is eye locationg in face. Method of face locating is using regiongrow on binarized image; Method of eye locating is using sobel operator to extract eyes' edge and then using projection to get eye's position.(2) Eye state judgement: The percentage of eye's width over height is used to judge eye state. Comparing the percentage 'm' with the threshold 'M', if m>20%*M, the eye state is open state; If m<=20%*M, the eye state is close state. The coefficient 20% comes from PERCLOS method.(3) Eye tracking: Kalman tracking method is used to estimate the eye's position in the following image, so the eye's searching range is decreased and real-time can be ensured.(4) Fatigue judgement: Based on PERCLOS method, caculating the number of images in which eyes are close state, denote n, and the number of all processed images, denote N. If n/N>=40%,this shows driver is weary.Experimental results proves the vality of methods in the software. At the same time, it is found that the software is sensitive to illumination, the results will be more accurate if use the monitor which is sensitive to infrared and use infrared irradiate the driver's face.