Research On Driving Behavior Evaluation System Based On Multi-sensor Information Fusion Technology

Recent studies have indicated that the probability of a traffic accident is particularly high,when it involves lane changing,turning or crossing.In the above situations,the driver needs to control the speed reasonably,turn on the indicate light and observe the traffic situation simultaneously.To rectify drivers’ non-standard driving behavior and reduce the incidence of traffic accidents,a real-time driving behavior evaluation system is proposed in this paper.The system evaluates driving behavior based on traffic information and vehicle attitude and other vehicle conditions information.To improve the validity and accuracy of the evaluation system,a multi-sensor information fusion technology is proposed in this paper,which integrates both the camera and the attitude sensor data.In order to detect lane changing effectively,we merge the information of the lane marker and the yaw angle of the vehicle.Moreover,the head pose is used to judge the driver’s intention,and we apply the kernel principal component analysis method to improve the accuracy of head pose classification.Furthermore,the head pose classification algorithm is optimized by using the correlation between face position and head pose,which reduces the computation complexity.Our experiments have shown that the evaluation system has a good effect on detecting the traffic scene and driving behavior.According to the data fusion scheme based on the traffic rules,the system achieves good performance in judging whether the driver’s driving behavior is reasonable or not.