基于轨迹数据的货车停车目的识别方法  

Truck Stop Purpose Identification Method Based on Trajectory Data

作　　者：姚雅慧 张戎[1,2] YAO Ya-hui;ZHANG Rong(Urban Mobility Institute,Ministry of Education,Tongji University,Shanghai 201804,China;Key Laboratory of Road and Traffic Engineering,Ministry of Education,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学城市交通研究院,上海201804 [2]同济大学道路与交通工程教育部重点实验室,上海201804

出　　处：《交通运输系统工程与信息》2023年第2期92-99,共8页Journal of Transportation Systems Engineering and Information Technology

基　　金：高端外国专家引进计划(G2022133042L)。

摘　　要：货车轨迹数据能够提供丰富的货运需求信息,然而从轨迹数据中识别停车目的的方法有待进一步研究。本文利用市内出行货车的轨迹数据,设计基于移动记录反推的货车停车点提取算法;提出融合货车出行调查数据的停车目的标定方法;构建包含停车点特征、出行特征、出行链特征、附近停车点特征和附近兴趣点特征这5个特征子集的货车行为特征集,作为识别模型输入变量;运用网格搜索算法,确定最优参数组合,建立基于二叉树支持向量机(SVM)的货车停车目的识别模型,区别货车装货点、卸货点和非装卸货点。以上海市内出行的重型货车为例,应用货车轨迹数据停车目的识别方法。结果表明:出车货车1 d平均停车4.6次;95%以上停车点的停车时间大于7 min,停车点间移动时间大于17 min;货车停车点提取算法的停车点个数误差约为6.5%;设定时间匹配标准和空间匹配标准是标定停车目的的可行方法;基于二叉树SVM的货车停车目的识别模型的准确率达85.83%。研究结果可为超大城市管理者分析城市货运需求、制定城市货运管理政策提供技术和数据支撑。The truck trajectory data can provide a wide range of information on freight demand,but the method for identifying stop purposes from the trajectory data needs further research.Using the trajectory data of intracity trucks,this paper designs a truck stop extraction algorithm based on reverse inference of movement records and proposed a stop-purpose calibration method incorporating data from the truck travel survey data.A feature set containing five feature subsets:stop features,trip features,trip chain features,nearby stop features,and nearby point of interest features is constructed and used as input variables for the identification model.A truck stop purpose identification model based on the binary tree Support Vector Machine(SVM)is developed using the grid search algorithm to determine the optimal combination of parameters.The model can distinguish between loading stops,unloading stops,and non-loading stops.Taking heavy goods vehicles traveling within Shanghai as an example,this paper applies the truck stop purpose identification method based on trajectory data.The results indicate that:trucks stop on average 4.6 times a day.Besides,more than 95%of stops have a stopping time greater than 7 minutes and a movement time greater than 17 minutes between two consecutive stops.The error in the number of stops of the truck stop extraction algorithm is about 6.5%.And setting the time-matching criteria and space-matching criteria is a feasible method to calibrate the stop purpose.Furthermore,the accuracy of the truck stop purpose identification model based on the binary tree SVM reaches 85.83%.This study could provide technical and data support for mega-city managers to analyze urban freight demand and formulate urban freight management policies.

关 键 词：城市交通 停车目的识别 二叉树支持向量机 轨迹数据 货车出行调查 

分 类 号：U491[交通运输工程—交通运输规划与管理]