Research On Detection Of The Driver Fatigue Based On The Fusion Of Multiple Features

As the development of the auto industry and the transportation industry, the safety of highway traffic has become a big problem which can not be ignored, for the traffic accidents have caused great loss in the property and damage to the society. Statistics show that driving fatigue is one of the main causes of the traffic accidents and so it is very necessary and meaningful to do researches on an effective method to detect the fatigue degree of the drivers.At present, some organizations have done some research about the detection of driver fatigue. There are also some detection systems under development and some simple detection equipments have been put into the market. But the accuracy of the detection can not meet the practical needs. More accurate and effective methods are still being developed.In this thesis, we analyze the principles brought out before which are used for kinds of detecting methods. After comparing the popular detecting methods abroad and researching their key techniques, we propose a fast detecting method of the drivers' degree of fatigue which is in the natural light, based on the computer vision and which doesn't need contact. The method uses a camera to capture the faces of the drivers in real time, extract the characteristics of their eyes and mouths which can best reflect their fatigue degree, and then uses the method of fuzzy logic to judge the physical characteristics to make an integrated decision rapidly and accurately.The main contents of this thesis are as follows:1) Do a detailed research on the key and difficult points of the present detecting methods and use a background upgrade algorithm based on dynamic matrix to realize the location and track of the head.2) Study the distribution of the color of skin, the relation of illumination and color spaces. Use a method combined with the adapted skin color segmentation and Adaboost algorithm to locate the face of the driver rapidly. 3) Study an eye location method with steps. First we use an accumulative histogram threshold method to binarize the face image. After the binarization, we use gray-level integration projection method to realize the rough location of the eyes. Then we get the exact position of eyes by template match method with a ring template. Finally we calculate blink frequency of eyes and the blink duration.4) Study a mouth location algorithm based on gray image. First, we get the general position of the mouth based on the pre-knowledge and the human facial organs distribution. And after enhancing the edges of the mouths, use the OTSU method to get the proper threshold to binarize the image. After that, we can get the mouth area from the image and decide whether the driver yawns.5) Analyze the current popular information fusing methods. Use the information fusing method based on fuzzy logic to realize the fusion of many physical characteristic parameters and then judge the fatigue degree of the drivers so as to improve the reliability of the detecting method.