A Driving Safety Assessment Method Based on Vehicle Trajectory Geometry

郭朴耀, 曾颖, Yan Wang, Yuying Li

Abstract: Drivers are autonomous and random in the process of vehicle operation, and their micro-driving behavior characteristics are related to traffic safety. Vehicle trajectory is the direct data that intuitively reflects the driver's micro-driving behavior, and drivers with different safety awareness and driving habits will generate different trajectory geometry. In this paper, through analyzing a large number of vehicle trajectory data to find drivers’ irregular or dangerous driving behavior, a driving safety assessment method based on vehicle trajectory geometry is proposed. Firstly, the invalid point of vehicle trajectory is eliminated using the K-Means algorithm. And then the characteristics of vehicle are analyzed based on the curvature change of trajectory geometry, the Hoff transformation and other methods to establish the trajectory structure with linear elements. Finally, a similarity method of vehicle trajectory is proposed to quantify a trajectories by linear elements, and evaluate the safety index of the trajectory, which can provide safety reference for regulating drivers’ driving habits.