Research Of Driver's Drowsy Detection System Based On Face Detection

In Modern times, drowsy driving is the one of the most severe factors caused traffic accident which brings serious problem and great effect to families and society. So in such bad environment, it is necessary and meaningful to study the driver's drowsy detection system to help the drivers drive safely. That system can play an very important role to reduce the traffic accident rate.That paper is mainly including four parts: Face preparatory detection based skin color; Face detection based dynamic filter and template matching; drowsy driving detection based on PER-NOFACE rule.At the face preparatory detection part, using the driver's face skin identity to position the skin region, as study indicate, the face skin distribute relatively centralized in color space. The color information could distinguish between face region and non-face region to some extent. In order to eliminate the impact caused by illumination condition, lighting compensation processing and nonlinearly transformation to YCbCr space were applied.Aiming at the problem that traditional skin-segment algorithm could not efficiently divide face from background whose color was similar to face's, a reliable face location algorithm based on dynamic ellipse filter was presented in this paper. A possible position and the size of a face were predicted through integration projection. In combination with Density of the ellipse, the filter would be adjusted to make itself best so that it could remove the region that are connecting to the face effectively. Experimental results show that the algorithm has good performance in face location.The second innovation in this paper is that a new fatigue detection algorithm was proposed which can use the area of face based on PER-NOFACE criteria to determine whether the driver is in the state of fatigue or not. This algorithm can overcome the weakness of HPM, whose complex equipment and appearance are not accepted by people; also solve the problem that traditional method PERCLOS cost large of computing capacity.Finally, the whole fatigue detection system was tested. The algorithms that using dynamic elliptical filter to eliminate the region whose color is familiar with skin color has good results and enhances PER-NOFACE criteria for testing the accuracy of fatigue. At the same time, PER-NOFACE is an innovative method which is different from other traditional criteria for fatigue testing. Compared to the traditional fatigue detection algorithm, despite a slight decline in accuracy, but it reduced computing capacity because of further test of state of eyes, mouth and improves real-time. This algorithm proposed has a certain practicality.In future work, how to combine a variety of means to improve the detection accuracy of the system and ensure that the substantial decline in computing volume and maintain real-time will at the same time will be the focus of the work. This is also fatigue detection system to develop a future direction.