【作者】 赵新勇；

【导师】 安实；

【作者基本信息】 哈尔滨工业大学 ， 交通运输规划与管理， 2013， 博士

【摘要】 高速公路网络的迅速建设和发展推动了我国的经济增长与社会发展，同时也消耗了大量的资源和能源，高速公路交通安全事故频发，严重影响人民生命财产安全，并对环境产生了巨大的影响。随着构建和谐、健康、可持续发展社会进程的逐步加快，高速公路交通安全预测和评价已越来越受到管理者的重视和研究者的关注。如何在多源异构交通安全数据的基础上，通过数据融合和数据挖掘分析影响交通安全水平的关键因素，并制定合理、可靠的交通管理和控制策略，降低交通事故人员伤亡率和经济损失，获得最佳的交通运输效益，已成为高速公路交通安全研究中迫切需要解决的实践难题和理论研究的前沿性课题。本文的研究遵循区域高速公路网络交通安全“多源异构数据融合、事故风险研判、安全等级综合评估、交通安全性能评价、案例应用”的思路，开展基于多源异构数据的高速公路交通安全性能评估方法研究。首先分析区域高速公路网络交通安全多源异构数据的来源，包括事故数据和环境数据两大类型；基于数据融合理论，剖析了高速公路交通安全数据串联、并联和混合三种融合结构，给出了高速公路交通安全数据融合系统的功能结构，建立了关联规则多维Apriori算法的流程，并给出了算法的描述，实现了高速公路事故多发点、路段和路网三个层次交通安全评估指标关联要素的甄别。其次结合支持向量机和决策树构建了基于交通流参数的高速公路交通安全事件风险研判技术。为了分析交通事故发生与交通流状况之间的关系，建立了基于支持向量机算法的高速公路交通事故风险研判模型，选择考虑支持向量机分类器和变量之间关系的特征变量选择方法，筛选出影响高速公路交通事故发生的显著特征变量。本文所构建的支持向量机预测器是基于相应时间组的最优特征变量以及核函数参数，同时比较了变量选择过程对事故风险预测结果的影响。再次，在多源异构数据融合处理的基础上，剖析了恶劣天气、不良交通状态等因素对事故成因的影响。综合考虑驾驶人反应时间、能见度、路面摩擦系数、坡度、车速、车间距等因素，提出了高速公路行车安全性综合评估模型，分析了自由流状态下、跟车状态下的行车安全性，并以算例验证了所提出的模型。然后，借鉴美国交通部采用的交通安全性能函数SPFs（Safety Perform-ance Functions）来描述和预测区域高速公路网络交通安全，构建符合我国区域高速公路发展实际的交通安全性能函数，给出了高速公路交通安全水平分析方法。提出了区域高速公路网络交通安全性能评价指标体系，将事故多发点、高速公路和路网交通安全事故整合在一起，从“点-线-面”的层次将多源数据结合起来，采用模糊区间理论进行综合评价。并以浙江省高速公路网络为例进行了具体的模型应用及分析。最后，以江苏省高速公路网络为案例，具体分析了江苏省公路网交通安全态势评估系统的应用情况。预测了江苏省高速公路追尾事故风险，评估了江苏省高速公路交通安全等级，评价了江苏省高速公路网络交通安全性能指标，结合评价结果提出了相应的交通安全改善建议。更多还原

【Abstract】 The rapid construction and development of freeway in China have promotedthe economic growth and social development, but also have consumed a lot ofnature resources and energy, caused traffic safety accident-prone, and had a hugeimpact on the people life and wealth, as well as the environment. With the estab-lishment of a healthy, harmonious, and sustainable development society, freewaytraffic safety has been paid more attention from managers and researchers.On the basis of multi-source heterogeneous traffic safety data, through datafusion and data mining to analyze the key factors affecting the level of safety,and develop a reasonable, reliable traffic management and control strategies toreduce the traffic accident casualties and economic losses, aiming to get the besttransportation efficiency, has become a cutting-edge topics in freeway trafficsafety researches. This thesis followed the technical route of regional freewaynetwork “multi-source heterogeneous data fusion, accident risk judgement, inte-grated assessment of safety rating, traffic safety performance evaluation, caseapplication”, based on multi-source heterogeneous data fusion to explore freewaytraffic safety performance evaluation.Firstly, the sources of regional freeway network traffic safety multi-sourceheterogeneous data were analyzed, including accident data and environmentaldata. Based on data fusion theory, three fusion structures of the freeway trafficsafety data were put forward, including the series, parallel and mixed fusionstructure. Then presented the freeway traffic safety data fusion system functionstructure, and established multi-dimensional association rules Apriori algorithmprocesses, and gave the description of the algorithm, which can achieve the indi-cation of freeway traffic safety evaluation elements, such as the accident-pronepoints, road level and road network level.Secondly with the combination of the support vector machine and decisiontree method, this chapter constructed the freeway traffic safety risk judgementtechnology through the traffic flow parameters. In order to analyze the relation-ship of traffic accidents and the conditions of traffic flow based on support vectormachine algorithm to establish the freeway traffic accident risk prediction model.Considering the relationship between the SVM classifier and variable feature se-lection method, to filter out the impact of traffic accidents occurred the salient features of variables. Based on the corresponding optimal feature variables andkernel function parameters to build the SVM predictor, and compared the vari-able selection process on the accident risk prediction results.Thirdly, based on multi-source heterogeneous data fusion, this chapter ex-plored the impact of inclement weather and inclement weather. Then proposedregional freeway network traffic safety situation assessment model under inclem-ent weather, especially the safely operation vehicle speed under different visibil-ity and precipitation conditions. And established the freeway traffic safety inte-grated assessment model to analyze the traffic safety under free flow state andcar-following state, a numerical example was proposed to verify the model.Then, referred to the SPFs （Safety Performance Functions） used by theU.S. Department of Transportation, to describe and predict the regional freewaynetwork traffic safety. This chapter established complied with the development ofChina’s regional freeway traffic safety performance function model, and pre-sented freeway traffic safety level method. And proposed regional freeway net-work traffic safety performance evaluation model based on fuzzy interval theory,which integrated the accident-prone points, freeways and road network traffic ac-cidents together, and from the "point-line-surface" level to combine the multi-source traffic data. Then this chapter applied this model in Zhejiang Provincefreeway network to analyze traffic safety situation.Finally, this paper took Jiangsu province freeway network as an example,analyzed the application of Jiangsu road network traffic safety situation assess-ment system. As well as predicted Jiangsu freeway rear-end accident risk, as-sessed the level of freeway traffic safety in Jiangsu Province, the traffic safetyperformance evaluation of Jiangsu Province freeway network, and according tothe evaluation result, this paper presented some recommendations for the im-provement of regional freeway network traffic safety.更多还原