Design And Research Of The Real-time Driver Fatigue Monitoring System Based On Video

Now, with the improvement of people's living standards in china, cars have already entered millions of households and become a major means of transport in people's daily out. However, our country's traffic accidents also increased rapidly. It became a social problem can not be ignored and a threat to people's life security as well as social order and stability. Fatigue driving, as one of the most important reasons for the increasing traffic accidents, the study of it is still in the beginning stage in our country and lag far behind the Western countries. At present, the research of the driver fatigue monitoring system based on video is still in the beginning stage in our country. So it is of great scientific value and significance.This paper based on analyzing of the domestic and international research status and development trend in the field of fatigue driving monitoring, summed up the current commonly used methods to monitor fatigue driving. Then a real-time driver fatigue monitoring system based on video is presented. This system evaluates driver's fatigue status by monitoring the changes of his eyes from video and is a much more realistic method.The real-time driver fatigue monitoring system based on video mainly includes three parts. They're the image acquisition module, the image processing module and the SVM classifier. The research and the design issues of this paper are the image processing module and the SVM classifier. The image processing module used the eye detection and tracking algorithm based on convolution kernel mask operation to deal with the video images of driver's face which gain from the image acquisition module. It is based on the Gabor transform to extract fifty-four multi-scale, multi-orientation Gabor features and the other two eye features. Then, these fifty-six features which associated with fatigue driving are put into the trained SVM classifier to identify whether the driver is fatigue and alarm in time. Thus, it can prevent traffic accidents from happening effectively.The major jobs of this paper are the following: 1. The paper summarizes detail the current domestic and international monitoring methods and major equipments of fatigue driving. Based on analysising the study status and existent questions of domestic fatigue driving monitoring deeply, the author proposes a effective program, which used an infrared camera's "bright pupil effect" to real-time monitor driver's eye charaters and evaluate whether the driver is fatigue. And designing the system's whole framework of hardware and software.2. Researching and designing the image processing module of the system. According to the real-time request of the system, the paper proposes the eye detection and tracking algorithm based on convolution kernel mask operation and designs the eye template based on "the bright pupil effect". Through the study of the eye detection and tracking algorithm, the paper design and implement this algorithm on the MATLAB.3. For the feature extraction unit of the image processing module, the paper analysis and studys of texture features, multi-scale transform and others image processing technologies, proposeing a kind of feature extraction method based on the Gabor transform. This method uses the eye's inside corner, outside corner, the center between the inside corner and the outsidecorner to the Gabor transform. For each point, take n =0,1,2 ,θ=0,(?),calculate its Gabor features. Then we can get a total of fifty-four Gabor features. Meanwhile, combining the other two features, which are the duration of eye closure and the average duration of the past ten times blinks as the feature vector of the image processing module's output.4. The paper also reseachs and designs the SVM classifier of the system. Through introducing and overviewing the theories and the basic conceptions related with the SVM classifier. It proposes the design method of the SVM classifier. Furthermore, using SVM OSU₃.0 toolbox on the MATLAB platform to train and test several groups of video clips, it achieves the desired results.The research work of this paper only take the eye features as feature parameters to evaluate whether the driver is fatigue. In the future study, it can also be combined with physiological signals, vehicles parameters and other indexes for the more comprehensive consideration. Embedded system can also perform experiment on and make products of it, expanding the field of its application.