Y. Yang, B. Sun, J.R. Pan

Pages: 87-100

Abstract
Aiming at the problems of low detection accuracy and long detection time in traditional driving behavior safety detection methods, a driving behavior safety detection method based on vehicle trajectory data is proposed. Firstly, vehicle trajectory data is obtained through LSTM neural network trajectory prediction model, and the obtained vehicle trajectory data is processed; Then, under the analysis of inherent characteristics of driving behavior, the characteristic parameters of risky driving behavior, including overspeed, rapid deceleration, and rapid acceleration, are determined through driving behavior speed and vehicle start and stop status; Finally, a strong classifier is constructed using an adaptive lifting algorithm, and the determined feature parameters are input into the strong classifier as raw data samples for driving behavior safety detection. The experimental results show that the proposed method has higher detection accuracy and shorter detection time for driving behavior safety.
Keywords: vehicle trajectory data; driving behavior; safety testing; LSTM neural network; adaptive lifting algorithm