城市快速路实时交通流运行安全主动风险评估

发布时间：2018-04-19 00:21

本文选题：城市快速路 + 交通事故　；参考：《同济大学学报(自然科学版)》2014年06期

【摘要】：基于上海市快速路系统采集的线圈检测器数据和事故数据,应用贝叶斯网络(BN)模型对快速路实时交通流参数与事故风险进行建模分析,并利用可有效应对缺失数据的高斯混合模型和最大期望算法分别对BN模型输入和参数进行估计,进而主动评估快速路实时交通流运行安全风险,并对事故状态提前做出预警.分别对可能事故点前后2组检测器和4个时间段的8组交通流数据进行了建模,结果表明使用事故发生地点上下游各一个检测器在事故发生前5~10min内的交通流数据建立的BN模型效果最好,事故预测准确率为76.94%.最后不仅与朴素贝叶斯分类、K近邻、反向传播(BP)神经网络等经典事故风险估计算法进行了对比分析,还与现有的实时风险评估成果进行了对比,结果表明BN模型预测效果最好.
[Abstract]:Based on the coil detector data and accident data collected by Shanghai Expressway system, the Bayesian Network (BN) model is used to model and analyze the real time traffic flow parameters and accident risk of expressway.Gao Si mixed model and maximum expectation algorithm which can deal with missing data are used to estimate the input and parameters of BN model respectively, and then evaluate the safety risk of real-time traffic flow on expressway actively, and warn the accident state in advance.Two sets of detectors and eight sets of traffic flow data of four time periods before and after the possible accident point were modeled, respectively.The results show that the BN model of traffic flow data in 5~10min before the accident is the best, and the accuracy of accident prediction is 76.94.Finally, compared with classical accident risk estimation algorithms, such as naive Bayesian classification and BP neural network, it is also compared with existing real-time risk assessment results. The results show that BN model has the best prediction effect.
【作者单位】：同济大学道路与交通工程教育部重点实验室;
【基金】：国家自然科学基金(51278362) 教育部新世纪人才计划(NCET-13-0425)
【分类号】：U491.1

【相似文献】