基于时空GPS轨迹的精细道路数据快速获取与变化检测

With the rapid urbanization and deteriorated traffic environment, they are in urgent need the precision and up-to-date road information when people are driving, and at the same time, they are producers and perceivers of fine road data themselves, because their driving trajectories can cover every lane and every road in the city road network. The famous specialist, professor in Carnegie Mellon University, Mitchell pointed out that “detailed road information can be mininged from vehicle trajectories” in the famous journal “Science” in 2010. However, the GPS on vehicles which is used to record driving trajectories is of limited positioning accuracy, and how to select high-accurate data from raw GPS trace and realize sophisticated road information mining and change detection at lane level is the problem that scientists face today. .This project studies high qualility trajectories road coverage detecting model with consideration of driving behavior and space time distribution characteristics of trajectories based on the mass trajectories continuously collected by GPS devices installed on taxis. The project solves the scientific problem of filtering available and accurate trajectories on the condition of low-accuracy GPS data, and studies trajectories clustering method based on the similarity of directed line, and solves the problem of mining sophisticated and multilevel road information from floating car data, and studies real time change detection of road topology based on the match of GPS trajectory vectors, breaking through the traditional road change detecting patterns and realizing a real-time collecting, low-cost, and fast updating approach of sophisticated road information mining and change detecting, which has important study value and application prospect today.

随着城市交通的飞速发展与日益复杂，人们出行时对精细道路数据需求迫在眉睫，同时也是精细道路数据产生者和感知者，国际顶级期刊《Science》2010 年发表论文“Mining Our Reality”指出“行车轨迹蕴含丰富道路信息”，如何从低精度车载GPS 轨迹中滤选出高精度数据，实现车道级精细道路信息挖掘提取与变化检测，是全球科学家面临的难题。..本课题基于出租车GPS 轨迹数据，研究顾及驾驶行为与轨迹时空分布特征的轨迹数据可信道路覆盖探测模型，解决低精度GPS 条件下高精度轨迹数据优选的科学难题；研究基于向量相似性的轨迹簇聚类方法，解决从GPS轨迹数据提取精细多级道路数据的难题；研究基于GPS 轨迹向量匹配的道路拓扑实时变化检测方法，突破传统变化检测模式，实现一种基于浮动车轨迹的采集易、成本低、更新快的精细道路信息获取与变化检测新方法，具有重要研究和应用价值。

高精度道路地图作为未来十大科技发展方向之一的无人驾驶最核心关键技术，成为Google、苹果、特斯拉、Here、四维图新、百度等国际科技巨头竞相争夺的技术战略制高点，长期面临着欧美发达国家以国土安全为由的“卡脖子”。针对现有地面测绘车、航空摄影测量、航天卫星遥感等专业测绘全球性道路数据“采集难”、信息“提取难”、模型“表达难”、变化“更新难”等问题，创建了车道级高精度道路众包测图新理论方法，突破轨迹清洗、车道更新等技术，提出多层次三维路网模型，研发众包测图系统，牵头研制国家标准，形成采集易、成本低、更新快、全球化应用的高精度道路众包测图新技术体系。.该项目形成了大数据环境下高精度道路众包测图成套技术，“足不出户，精制天下图”，出版专著1部，在交通、GIS、制图等国际顶级期刊IEEE ITS、TRC、IJGIS、CAGIS等发表论文41余篇（18篇SCI，20篇EI），授权专利9项，2017年以刘先林院士为组长的技术鉴定委员会一致认为“达到国际先进水平”，获IEEE会士等国际同行高度评价，获2019年教育部科技进步一等奖（排名1），2017年度国家测绘科技进步一等奖（排名1）。项目成果在华为、四维图新等国际知名企业广泛应用，支撑智慧城市、智能交通、无人驾驶等全球前沿技术发展，打破欧美“卡脖子”封锁，填补我国道路众包测绘的技术空白，实现道路测图从专业测绘向众包测绘的技术跨越，成为专业道路测绘的重要补充，是我国智能交通与测绘科技进步的划时代标志性成果。