中文摘要：

针对大规模浮动车数据量庞大、地图匹配时对实时性要求高的特点,提出一种离线计算与在线匹配相结合的浮动车实时局部地图匹配方法.在离线计算部分,首先对对矢量道路网进行二级网格化分,然后对二级网格的每一个栅格,以其中心点坐标到各个候选路段或道路节点的距离以及候选路段的通行方向计算距离和方向的组合相似度,根据组合相似度最大值与次大值之差的大小分成4种情形,并将组合相似度明显占优的候选路段或者道路节点作为该栅格的属性值预先存储;在线匹配时,针对实时获取的轨迹点,采用文中提出的“双向前行-后行法”快速判定前后时刻欲匹配路段之间的路径可达度,并据此判断所属情形;当满足条件时,可立即获得当前轨迹点的匹配结果,即该轨迹点所在栅格的属性值.实验结果表明了该文方法处理海量浮动车数据的可行性和有效性：通过对深圳市大规模浮动车实测数据进行验证,该方法比局部地图匹配方法的单点匹配平均耗时减少了1倍;对整个深圳市区构建两级网格并存储属性值,其占有存储空间约为8 MB,可以将其一次性读入内存;整个深圳市道路网的离线计算时间约需花费32h,但离线计算一经完成就可以长期服务于实时匹配.

英文摘要：

Aiming at the demand on fast and real time map matching for massive floating car, this paper proposed a local map matching algorithm which combined online map matching with offline calculation. The Offline calculation firstly build two levels of grids for vector road network, and then combined similarity could be pre-caleulated offline which takes both the distance from center of each cell to every candidate matching road segment or node and traffic direction of every candidate matching road segment into account. According to the value of the difference between the maximal combined similarity and the second largest combined similarity, the related road segment or road node whose combined similarity clearly dominated could be stored in advance as attribute value of the cell in second level grid. When online map matching is implemented for every real time GPS point, the cell in the second level of grid where the GPS point fall within could be firstly instantly got, and then measure of path accessibility from previous GPS point to current GPS point could be quickly determined by means of two-way forward and backward method, and then final matching result of real time GPS point that is the attribute value stored in advance could be instantly obtained according to the value of path accessibility. Experiments on massive floating car in Shenzhen city show that the matching speed of this method is twice as fast as the matching speed of local map matching method, and it only occupied about 8 M storage space on storing two level of grids and its attribute values which could be read into memory once, and it took about 32 hours to finish offline calculation for the whole Shenzhen road network, however, once offline calculation is finished, it was able to provide a long history service to online map match.