Research On Influencing Factors Of Elderly Pedestrian Traffic Accidents Considering Built Environment

The main travel mode of the elderly group is walking,but with the physical inconvenience caused by the growth of the elderly group,the reaction speed of the elderly decreases,and they are more vulnerable to serious trauma and even death in traffic accidents.In recent years,the modeling of elderly pedestrian traffic accidents from both macro and micro aspects is one of the research directions of road traffic safety.The influencing factors of accidents are related to the complex built environment.Considering the spatial heterogeneity at the macro scale,thesis quantitatively analyzes the impact of urban environmental factors on the number of elderly pedestrian traffic accidents in the administrative Street area,further solves the problem of too discrete micro scale pedestrian environment data by establishing a data mining model,quantitatively analyzes the causes of the severity of elderly pedestrian traffic accidents,and provides technical support for the fine management of urban traffic.First,master the characteristics of elderly pedestrian traffic accidents.Based on the traffic accident data of Beijing in 2015,the accident details of elderly pedestrians and pedestrian groups are screened,and the accident characteristics are extracted and counted through five factors of time and space,personnel,vehicles,roads and environment.Compared with pedestrian group traffic accidents,there are significant differences in the occurrence area and location,gender,accident type,environment,number and severity of accidents.Then,the influencing factor model of the number of elderly pedestrian traffic accidents considering the macro built environment is constructed.The spatial autocorrelation of elderly pedestrian traffic accidents divided by administrative streets is tested based on Moran index.The common point and common edge regression matrix is constructed as the spatial weight matrix.The collinear variables are determined by Vif test and regressed step by step.The goodness of fit is compared based on OLS model and geographic weighted regression model,and the influencing factors of spatial heterogeneity are considered.The results show that the accidents are greatly affected by the regional average speed,the average distance between intersections,the density of elderly population,the density of road network and the density of service facilities,and show different differences with the distribution of geographical space.Finally,the cause model and prediction method of the severity of elderly pedestrian traffic accidents considering the micro built environment are constructed.Using the web crawler method,the street scene pictures of the location of elderly pedestrian traffic accidents in Beijing in 2015 are crawled,and the elderly pedestrian traffic accident scenes are identified based on two deep learning algorithms:semantic segmentation（Seg Net）and object detection（Yolo）,which are transformed into unstructured discrete data to obtain environmental parameters.In the modeling and analysis,based on Apriori association rule theory,mining the important influencing factors affecting the severity of elderly pedestrian traffic accidents.Quantitative analysis based on CNN neural network.The results show that the error of CNN neural network prediction model considering important influencing factors is 8.10430)^-5.Among the influencing factors,the vehicle speed of 25-50 km/h is more prone to accidents and the accident severity is higher;Accidents are more likely to occur in general road sections,and the severity of entrance and exit accidents is higher;Accidents are more likely to occur in sunny days,mainly in places with poor walking safety,comfort and convenience.Accidents in snowy days are more serious,mainly in places with poor walking safety,comfort and convenience.