基于多源特征融合的高速公路路网交通状态判别  

Traffic status discrimination of expressway network based on multi-source feature fusion

作　　者：邱暾 徐丽丽 王宇飞 苗英辉 杨杰 QIU Tun;XU Lili;WANG Yufei;MIAO Yinghui;YANG Jie(Liaoning Provincial Transportation Investment Group ATS Technology Co.,Ltd.,Shenyang 110166,China;Southeast University,Nanjing 210096,China)

机构地区：[1]辽宁交投艾特斯技术股份有限公司,辽宁沈阳110166 [2]东南大学,江苏南京210096

出　　处：《交通科学与工程》2024年第6期135-142,共8页Journal of Transport Science and Engineering

基　　金：流程工业综合自动化国家重点实验室(东北大学)开放基金(2023-kfkt-01)。

摘　　要：为满足管理者和出行者全面掌握高速公路路网交通状态的需求,以高速公路路网数据和收费数据为基础,采用多源特征融合的方式构建高速公路路网交通状态判别模型。首先,将高速公路路网划分为路段,提取路段的静态特征和动态特征,以多源特征融合的方式全面刻画路段特征;然后,采用自定义规则结合人工标注的半自动化数据标注方式对数据进行标注,保证原始数据标注的准确性;最后,针对数据不均衡的问题,提出由两层随机森林模型组成的双随机森林模型,利用第二层随机森林模型提升第一层随机森林模型分类的准确性。为了证明所提方法的有效性和准确性,收集了辽宁省高速公路的路网数据和收费数据对模型的效果进行验证,结果显示:畅通状态的准确率为100.0%,基本畅通状态的准确率为94.0%,轻度拥堵状态的准确率为98.0%,中度拥堵状态的准确率为94.0%,重度拥堵状态的准确率为97.0%,全网交通状态的平均准确率为96.6%。To meet the needs of managers and travelers to comprehensively grasp the traffic status of the expressway network,this paper employed expressway network data and toll data as the basis to build a traffic status discrimination model of the expressway network based on multi-source feature fusion,which was adopted for discriminating the traffic status of the entire expressway network.Firstly,the expressway network was divided into sections,and static features and dynamic features of the sections were extracted to comprehensively characterize road section features based on multi-source feature fusion.Then,a semi-automated data annotation method based on custom rules combined with manual annotation was utilized for data annotation to ensure the accuracy of the original data annotation.Finally,given the data imbalance,a dual random forest model consisting of two layers of single random forest models was proposed,with the second layer of the random forest model employed to improve the classification accuracy of the first layer of the random forest model.To demonstrate the effectiveness and accuracy of the proposed method,this paper validated the model by collecting expressway network data and toll data of Liaoning Province.The results show that the accuracy of the smooth status is 100.0%,the accuracy of the basic smooth status is 94.0%,and the accuracy of mild congestion is 98.0%.Additionally,the accuracy of moderate congestion is 94.0%,the accuracy of severe congestion is 97.0%,and the average accuracy of the traffic status is 96.6%.

关 键 词：交通工程 交通状态 状态判别 多源特征融合 收费数据 

分 类 号：U495[交通运输工程—交通运输规划与管理]