Combined Bayesian Network-Based Recognition of Lane Changing Behavior

Aiming at the lane change behavior recognition requirements of lane change warning system, natural lane change samples were captured by using a test vehicle. Steering angle and distance between vehicle and lane mark were used as characteristic parameters of lane change behavior. Support vector machine (SVM) method was used to establish recognizing model of lane change. According to the high-low identification accuracy of the different time window, 1.2 seconds was selected to be as the window length .The sample data in each time window were filtered by Kalman filter. Then, by using principal component analysis method, the first and second principal components were extracted from all principal components to be as the new variable. Support vector machine-bayesian filter identification model was built to identify general lane change behavior. Final recognition results show that the recognition rate for the real lane change samples can reach more than 95% and the proposed model can also meet the real time and reliability requirements of lane change warning system.