基于深度交通事故调查的驾驶人应急行为数据库建设及数据挖掘研究

项目名称： 基于深度交通事故调查的驾驶人应急行为数据库建设及数据挖掘研究

项目编号： No.51205119

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 机械工程学科

项目作者： 孔春玉

作者单位： 广东技术师范学院

项目金额： 24万元

中文摘要： 驾驶行为研究是世界智能汽车和智能交通领域的热点，缺乏真实的事故数据是严重制约我国深入研究的瓶颈之一。本研究的目的是基于真实的事故数据构建我国驾驶人应急行为数据库,并深入挖掘应急驾驶行为与道路弱势群体（行人、骑两轮车者）损伤风险的关系。数据来源于长沙市深度交通事故调查研究，通过问卷调查、事故重建和虚拟驾驶试验等方法采集驾驶人的真实应急避险行为数据；对感知、决策、操控等应急行为环节以及车辆运动参数、道路环境指标等进行精细化描述，提取应急行为特征值，构建我国乘用车驾驶人应急行为表征数据库。基于深入的事故形态研究和贝叶斯信念网络方法深入挖掘应急驾驶行为特征与弱势群体损伤风险的定性关系；基于Logistic回归方法探索应急驾驶行为特征与弱势群体损伤风险的定量关系；基于网络层次分析法（ANP）构建驾驶人应急行为风险评估模型。为我国车辆主动安全、人机工程以及智能交通技术的发展提供基础数据和科学依据。

中文关键词： 应急驾驶行为；数据挖掘；自行车事故；电动两轮车事故；贝叶斯网络模型

英文摘要： Driving behavior research has become one of the hottest spot in the researches on intelligent car and intelligent transportation system. However the lack of real accident data is the bottleneck of making an in-depth study. The current study aims to build the database of driver urgent behavior in China based on the real accident cases, and make an in-depth data mining with the relationship between driver urgent behavior and vulnerable road users' (pedestrian and cyclist) injury risk. The accident data were collected from in-depth traffic accident investigation in Changsha. Through the methods of questionnaire investigation, accident reconstruction and virtual driving test, the data of the driver behavior in urgency will be collected. In order to build the database of urgent behavioral characteristics for passenger cars' drivers, decision-making, controlling, and kinematic parameters of passenger cars, and indicators of road environment. The qualitative and quantitative relationships of driver urgent behavior and traffic injury risks of vulnerable road users will be studied based on the accident scenario analyses, Bayesian Networks and logistic regression method. The driver urgent behavior assessment model will be built based on Analytic Network Process (ANP). This study will provide the essential data and scienti

英文关键词： driver urgent behavior；data mining；Bicycle Accident;；Electric two-wheeler accident;；BNT model;