Research Of Driver Fatigue Monitoring System Based On Omnidirectional Vision

Traffic accidents are a serious problem all around the world, of which driver fatigue is one of the main causes. Therefore, studying driver fatigue and developing an effective driver fatigue monitoring system makes great practical sense.In this paper, we study and design a kind of driver fatigue monitoring system based on omnidirectional vision. First uses omnidirectional vision sensor to get the panoramic image, then gets three perspective images respectively containing the diver's face, the steering wheel and the road in front of the driver, by using perspective unwrapping algorithm, and then monitor three objects which can reflect the driver's fatigue, by using computer vision technology. The three objects are driver, steering wheel and the road in front of the driver, and among three objects, driver monitoring includes eye, mouth and head monitoring. Finally judge the driver' fatigue condition according to the information acquired by the above monitoring.The main work is summarized as follows:(1) Study the application of omnidirectional vision sensor and the unwrapping technology of omnidirectional image in driver fatigue monitoring. Choose an omnidirectional vision sensor which adopts wide dynamic range technology, and study the best way by which the sensor can be placed in the car, to make the the sensor to capture all monitored objects most clearly. Choose a suitable unwrapping algorithm for our system, and study the setting of various parameters of the unwrapping algorithm, to unwrap the omnidirectional image to perspective image that fits the visual habit better and is more suitable for analysis.(2) Study face detection and tracking algorithm. By analyzing and comparing various existing algorithm, finally adopt an algorithm mainly based on skin color. Taking into account that the driver's face is almost still, so design a simple quick tracking algorithm.(3) Study the monitoring algorithm of 5 objects which can reflect the driver fatigue, the 5 objects are eye, mouth, head, steering wheel and the road in front of the driver. Analyze and compare the existing commonly used algorithm, and search for most suitable algorithm for the system. Design a new algorithm in the monitoring of mouse and the road in front of the driver.(4) Develop a driver fatigue monitoring system based on omnidirectional vision sensor. The system uses Java language, and also calls some C++ language components. The system is developed by using component-based development model. The basic components designed by us can not only be applied to our system, but also can be applied to other image processing system.