The Research On Method Of Fatigue Detection Based On Facial Feature Fusion

With the rapid development of automobile industry and highway construction, the road traffic safety problems have become more serious. Due to the traffic accidents, huge economic losses to the state and society would be caused, people’s physical and psychological damage would be made. It is shown that driver fatigue is one of the main causes of accidents, therefore, it is necessary to test the driver’s fatigue. In this paper, the state of driver fatigue based on the corresponding visual algorithms and digital image processing is studied, by using the driver video recorded through a stationary camera.In this paper, the driver fatigue detection system was studied, according to the function of the system, the framework and overall system processes were designed, and the corresponding algorithms also were designed for each part of the system. Driver fatigue detection system was consisted by face detection, feature extraction and fatigue judgment. First, the collected images should be preprocessed, such as graying, denoising, illumination compensation et al. Then the driver’s head movement area was extracted by using background subtraction. In the face detection part,afterthe usual detection methods were analysed and researched, AdaBoost algorithm based on Haar-like features was used to detect face, where the classifier was gained by training sample,and the efficiency and accuracy of detection were improved. In terms of feature extraction, the eye state extraction and the mouth state extraction were introduced respectively. When the eyes were extracted, several common methods were introduced, then the method was proposed which do binarization to the image of the upper part face firstly, and then extracts eye feature using integral projection and template matching method, which can extract eye features quickly and accurately. When the mouth was extracted, the edge of mouth was enhanced by edge detection firstly, and then mouth features were extracted by binarization; In terms of fatigue judging, the methods using eye features to judge fatigue and using mouth features to judge fatigue were introduced firstly, and then the method which judged fatigue by combining the two features was proposed.In this paper, all of the above algorithms are experimented by using the VC++. It is shown that the feasibility and the efficiency of the algorithms have been ensured and the accuracy is improved by using the fusion method.