基于GPS的出行者停止语义推断模型  

Asemantic inference model for stop of traveler based on GPS

作　　者：刘春[1] 曹凯[1] 

机构地区：[1]山东理工大学交通与车辆工程学院,山东淄博255049

出　　处：《山东理工大学学报（自然科学版）》2015年第4期48-52,共5页Journal of Shandong University of Technology:Natural Science Edition

基　　金：国家自然科学基金资助项目(61074140);山东省自然科学基金资助项目(ZR2010FM007)

摘　　要：传统的车辆运动轨迹特性分析通常采用对其停止点加注语义的方法,该方法存在信息采集繁杂、信息遗漏以及实时处理不方便等问题.为此,提出以GPS轨迹数据中的语义信息为直接挖掘对象,运用马尔可夫链方法推断车辆运动下一停止点的方法.该方法通过辨识车辆停留中的子停留,利用车辆运动轨迹点特征参数(即时速度、停留时长等)进行信息量化,并与构建的判别信息库进行比对,从而挖掘车辆停留中的子停留语义信息.此外,通过划分车辆运行状态层次,统计每个子停留在每个状态层次的出现频率,以此推断车辆的未来停止点.实车测试结果表明,采集数据量越大,状态层次划分的越细,计算结果越稳定;预测范围越小,按照比例还原的电子地图产生的误差越小,所得的预测越准确.Traditional methods for vehicle trajectory characteristic analysis usually adopt filling the semantic information for the stop point.In this way,many problems have been caused such as making information collection complicate,leaving out information and making inconvenience in real-time processing.Aiming at this problem,on the basis of the above method,this paper puts forward the method of using Markov chain to infer the next stop point of the moving vehicle,based on mining the semantic information hiding in the GPS trajectory data directly.In this method,sub-stops are identified.The characteristic parameters（instant speed,the stay time,etc.）of the points in the vehicle moving trajectory are analyzed.Then,the above information is compared with the k-base to dig the semantic information of the sub-stops.In addition,the next stop point of the moving vehicle is inferred by dividing the vehicle running state level and counting the frequency of the sub-stops in each state level.Real vehicle test results show that the stability of the calculation results is improved by collecting more test data and dicing more state levels.The smaller the predicted range is,the less errors are produced when restoring the electronic map according to the percentage.

关 键 词：车辆运动轨迹分析 语义注释 子停留 马尔可夫链 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]