A method for identifying frequent points of electric vehicle and human non-collision accidents considering space-time characteristics

Abstract

This paper proposes a method for identifying high-incidence points for non-collision accidents between electric vehicles and pedestrians that considers spatiotemporal characteristics. The method obtains data on non-collision accidents between electric vehicles and pedestrians and road network distribution data; preprocesses the accident data; quantitatively analyzes the temporal distribution characteristics of accidents from the perspectives of monthly and hourly distribution; introduces a comprehensive impact index for category accidents, uses the analytic hierarchy process to determine the weights of accident influencing factors, and reveals the spatial clustering of non-collision accidents between electric vehicles and pedestrians through the weighted network kernel density estimation method; uses a density peak clustering algorithm as an accident spatial clustering model based on spatiotemporal characteristics; and introduces time dimension characteristics to construct an ST-DBSCAN model for identifying high-incidence points and segments in spatiotemporal accident locations. This method avoids the problems of large accident clustering areas, low dispersion, and clustering trends in some areas. It accurately identifies high-separation high-incidence points and segments in spatiotemporal accident locations, as well as high-separation high-incidence points and segments in spatiotemporal accident locations near a specific time point.

Images