Research And Implementation Of Driver Fatigue Detection Based On Video

With the rapid development of automobile industry, traffic safety issues are also increasingly prominent. A lot of statistical data shows that the fatigue driving is a major cause of the traffic accidents. So the research on detection of fatigue driving has a very important and practical significance. In the present study, the non-contact fatigue detection method based on machine vision have got good achievements. In this paper, based on state of the human eye in the video, a fatigue detection system is designed and implemented with the characteristics of drivers’ eyes state. In this paper, the major work and innovations are shown as follows:1. Study and analyse the methods of human face detection, and focus on the implementation and comparasion between the detection method based on complexion feature and Adaboost algorithm, the results show that the Adaboost algorithm is better on the detection error rate level,, on the basis of the realization of face detection, the Adaboost algorithm is used again to realize the human eye detection, and a method based on adaptive threshold value is used to exclude the non-human eyes in mistakenly identified areas.2. Study and compare the different filtering methods such as median filtering, the method of median filtering is chosen by the experiment of driving environment, and on the basis of the previous analysis, the different threshold segmentation principle are studied and analysed, the OTSU dynamic threshold algorithm is selected according to the actual effect.3. Study and compare the applicable features of different target tracking algorithm, the combination of kalman filtering algorithm and Adaboost algorithm, the human face tracking is implemented, which makes up the time wasting defect of the Adboost algorithm and speeds up the system testing time.4. Implement a driver fatigue detection system, which firstly positions the driver’s facial area by the video image, and then positions the human eye in the facial area. Durng the recognition of the eyes state, a recognized algorithm with the combination of ratio between the height and width of the eyes and the external rectangular area is used. Finally, an algorithm video of driver fatigue detection which combines PERCLOS and blinking rate is proposed to detect the driver’s fatigue state.