Identification of Dangerous Driving Behaviors Based on Neural Network and Bayesian Filter

Abstract

Identification and amendment dangerous driving behavior timely and accurately is a necessary means to reduce traffic accidents. This paper proposed a dangerous driving behavior identification method based on neural network and Bayesian filter. By using vehicle-mounted radars and cameras obtain movement state information of the vehicles around the host vehicle and lane line distance data, on the basis of which, the identification model is established. Then evaluate model performance by the real data. The test results show that after the correction of neural network output by Bayesian filter, the model accuracy has a sharp rise.