交通运输工程与信息学报2025，Vol.23Issue(2)：110-121,12.DOI:10.19961/j.cnki.1672-4747.2024.12.005
基于多源信息融合的城市快速路驾驶风险辨识与方法
Risk identification and method of urban-expressway driving based on multisource information fusion
摘要
Abstract
[Background]Owing to the rapid growth of car ownership,urban traffic congestion and safety problems are becoming increasingly severe.Urban expressways have become common loca-tions of traffic accidents,and traffic-accident rates constitute more than 50%of urban-highway traf-fic accidents.[Objective]To analyze the interaction and correlation mechanism between driver-state change,driving style,and driving-risk degree in urban expressways,four driving-risk identification models based on multisource information fusion are proposed,the driving danger levels of drivers with different styles are determined,and an optimal model is identified.[Methods]Based on the physiological,psychological,and vehicle operation data of a driver,a driving dataset with multi-source information fusion is constructed.First,the driver's driving style is obtained via Pearson cor-relation analysis and K-means clustering.Second,the entropy-weight method and driving-style la-bels are used to quantify the driving-risk rates of drivers with different driving styles.Finally,MLP,SVM,CNN,and LSTM are used to develop four driving-risk identification models,and a model-per-formance evaluation index is used to select the optimal model.[Results]Based on the vehicle opera-tion data,the driver's driving style is classified into conservative,stable,and impulsive type,and the driver's risk level is quantified into three categories,i.e.,low,medium,and high,based on the driv-ing-style results.Among the four driving-risk identification models,the CNN-LSTM model demon-strates the best recognition effect,with an accuracy rate of 0.90,which can effectively identify the driver's risk level.The average change rate of the driver's speed at the entrance and exit of an urban expressway ramp is 22.14%and 14.8%,respectively.[Applications]The results of this study pro-vide a reference for managing urban-road traffic safety and preventing human-induced road acci-dents.They enable targeted management of high-risk drivers.Future studies should incorporate eye-movement data to further enhance urban-road safety.关键词
城市交通/驾驶风险/深度学习/多源信息/驾驶风格Key words
urban transportation/driving risk/deep learning/multisource information/driving style分类
交通运输引用本文复制引用
朱兴林,陈梦瑶,刘泓君,王光东,姚亮..基于多源信息融合的城市快速路驾驶风险辨识与方法[J].交通运输工程与信息学报,2025,23(2):110-121,12.基金项目
2024年新疆维吾尔自治区自然科学基金项目(2524KJTZRJJ) （2524KJTZRJJ）
2024年新疆维吾尔自治区高校科研计划项目(2224GXKYJH) （2224GXKYJH）
数字时代新疆高校交通运输专业学位研究生优质课程建设与实践项目(XJ2024GY14) （XJ2024GY14）
