The Basic Research Of Detection Of The Driver Unsafe Behavior Based On Multiple Features

Since China joined the world trade organization,the economy and national infrastructure construction has improved.The number of private cars is increasing year by year,traffic accidents occur frequently and have a significant impact on lives and property of people.Therefor to develop the system of supervising the behavior of drivers to enhance the awareness of safety and skill of driving has high practical value and social significance.Aiming at the driver assistance system based on computer vision is mainly focused on the mental state and fatigue driving of drivers,but the behavior of drivers have impact on safety of driving,so this paper takes the driving behavior of drivers as the breakthrough point to test the unsafe driving behavior.The main research content are as follows:(1)This paper divide the hand behavior of drivers into two categories: the hand movements trends of drivers and the position of hands on the steering wheel.The system get the hand movements of drivers through the steering information and gear information of vehicle,and use the computer vision to locate the hand position of drivers.(2)To separate the hand of drivers from complex background,this paper use skin color model.In order to make the skin color model cope with different external environment,an adaptive skin color model is put forward.(3)Taking advantage of the collection of driver behavior characteristics,we train a decision tree for driving behavior recognition,finally we use a lot of samples to validate the feasibility of this method.