Driver Fatigue Detect Based On Fusing The Eyes And Mouth Features

The research of driver fatigue detection and early warning approach has attracted wide attention. The driver fatigue detection method is studied by many of research institutions at home and abroad, related research has made many achievements after years of development, and there are some products input to applications. However, testing and practical experience has shown that the current driver fatigue detection methods are still limited in practice:such as the low measurement accuracy, more restricted scenarios, weak robustness, and difficulties in achieving real-time monitoring and so on. Therefore, the study of a more practical approach of the driver fatigue detection and early warning has great significance.To improve the accuracy of the fatigue detection, this paper is to be through integrated the fatigue-related features such as the eye and mouths and so on. In order to round-the-clock monitoring and analysis the status of the driver, this paper is to through infrared camera for infrared images in real time for the driver, and then extract the visual characteristics of the eye and mouth, finally this paper is to be through analyzed of the eyes status of opening and closing and the mouth yawning or not, for analyzing and determine the driver fatigue status. The main work of this paper is as follows:(1) Improved facial features of eyes and mouth point positioning strategy. For the problems of low detection rate and high error rate of directly detecting the eyes and mouth feature points from the video images, in this paper, first presents the classical AdaBoost algorithm to locate the face region, and then to locate the regions of other related features in the region of face.(2) Improved the search strategy of face detection. To achieve the purpose of real-time face detecting, combined with the driver under normal circumstances, the head relatively small range of features, this positioning of the previous frame to the next frame the face position estimation about the location of face to narrow the search to improve the human face Detection efficiency.(3) For the problem of the system's high false alarm rate caused by the driver's habits of sleeping with the eyes opening and frequent blinking. In this paper, we joint PERCLOS, blink frequency, blink duration of the eyes fatigue parameters for fatigue judging. When the driver has the habits of sleeping with the opening, PERCLOS and blinking frequency are relatively small; if the driver's habit is frequent blinking, their blink rate is high, blink duration is short.The simulation results show that the proposed has a certain improve in the system's real-time and accuracy. This article builds demo driver fatigue detection warning system by using Visual Studio 2005, it not only of the basic experimental platform of follow-up study, but also of the foundation for developing the application software in the future.