Research On Vehicle Driving Behavior Recognition Based On Vehicle Video

With the improvement of the quality of life,many people have purchased cars as a means of travel,followed by an increase in traffic accidents.It is a tedious and time-consuming task to analyze the vehicle driving behavior by looking up the vehicle video manually.Therefore,it has a strong application value to analyze the vehicle driving behavior automatically by using machine learning method.The research contents of this paper can be divided into three parts:First,the lane image enhancement method used to identify vehicle driving behavior is studied.For the lane image whose average brightness is greater than the threshold brightness,the dark primary color principle defog algorithm is directly used to enhance the clarity of lane image.For the lane image whose average brightness is less than the threshold brightness,the Msrcr algorithm is first used to enhance the brightness of the image,and then the dark primary color principle defog algorithm is used to enhance the clarity of the image.The experimental results show that the method can make the lane line clearer,and the lane line in the dark area can be well displayed.Secondly,a lane fitting,tracing and lane type recognition method for vehicle driving behavior recognition is proposed.In order to extract lane lines from Lane images more effectively,the lane images are thinned,and the lane lines are fitted by the combination of cyclic Hough and least squares.The extracted lane lines are tracked by Kalman filter,and the dynamic region of interest is established to make the lane detection more stable and reduce the amount of data processed.Finally,based on the edge and color features of the lane line,the white dot proportion method is used to identify the type of lane line.The lane line is divided into white solid line,white dashed line,yellow solid line and yellow dashed line.The experimental results show that the lane line identified by the method accords with the actual Lane type.Thirdly,a method of vehicle behavior recognition is proposed.Firstly,the lane angle of the lane where the vehicle is located is used to define the offset,and then the vehicle's offset behavior is identified.On this basis,the vehicle's lane change behavior is identified according to the continuous tracking of the offset markers.When the lane change occurs,the violation is judged according to the lane type.The experimental results show that the proposed method can effectively identify the vehicle's offset and diversion behavior,and the method is simple and easy to implement.