Research On The Inspecting Technique For Driving Fatigue

Driver fatigue is one of the important causes in the frequent traffic accidents at present. Therefore, it is significant to monitor and prevent driver fatigue efficiently for reducing the traffic accidents.In this thesis, the current methods of the detection on driver fatigue are contrasted. The key problems and difficulties of these technologies are analyzed. Then two detection systems for driver fatigue are established——natural light and infrared light. We lay a video camera to film the driver's face. Eye features, which are cues for determining driver's state, are extracted from the video of face. The main task in this paper presents as follows:(1) In constructing drowsy status recognition system, the necessary image pretreatment technique is investigated. Natural light: Neighborhood mean filtering method is used to eliminate the random noise in the input images immediately. The operation of the method is simple. The light compensation is achieved to reduce the color deviation with "reference white" method. Infrared light: Median filtering is used to eliminate the isolated noise in the video image. Edge information is protected with the method and the method is good for edge distill.(2) The arithmetic of human face detection is improved on. Natural light: Based on the good clustering character in the color space of skin color, a clustering elliptical model based on the YCbCr color space is set up. The color model is used to transform color images into binary images, to separate the skin color from non-skin color regions. The face skin identity is highly adaptive to the geometry variation such as face zoom and face circumrotation. Because of the fast computing and simple course, skin information arithmetic meet the demand of real time. Infrared light: The Otsu adaptive threshold method is used to transform infrared images into binary images, to distinguish between human face and background. Improved integral projection algorithm is used to mark the location of human face. Compared with other traditional algorithms, the improved integral projection algorithm lessens the region of human face and makes much more progress in real time.(3) Kalman filtering algorithm, which can improve the eye tracking speed and accuracy, is used to predict and track of the face region.(4) A new algorithm on eye location and detection is designed. Natural light: A combination method of integral projection and mathematical morphology to locate and monitor the eyes is adopted. Infrared light: A combination method of sobel edge detection and hough transform to locate and monitor the eyes is adopted because of the eye characteristics at special wavelength. Finally, PERCLOS parameters are calculated to determine driver fatigue.The character and applicability of these two methods are discussed in detail. A great deal of experiments are done, which show that human's eyes and eye fatigue are accurately monitored with these two methods and these two methods are practical and reliable.