隐马尔可夫模型路网匹配的MapReduce实现

MAP MATCHING WITH HIDDEN MARKOV MODEL ON MAPREDUCE

离线路网匹配考虑全局数据以求更高匹配准确性,因为地图的分割放弃对路网连通性这一严重影响匹配准确性的特性的考量,已有的一些研究尝试使用分布式计算框架来实现。另一方面,以往的研究局限于较小的地图空间,缺乏针对路网匹配应用下大规模地图数据的有效管理。首先提出分布式隐马尔可夫路网匹配模型。该模型利用维特比算法的可分割性,将隐马尔可夫模型这个准确性很高的路网匹配算法运行在分布式计算框架上,在不丢失匹配准确性的前提下,实现了在将路网匹配根据地理位置分而治之的同时减少地图I/O操作,提升计算效率达到50%。

Off-line map matching pursues higher matching accuracy in view of the whole trajectory. There were studies trying to use distributed computation framework to realize the research since the map segmentation fails to achieve road connectivity,which slashes the consideration of matching accuracy. On the other hand,the former researches were limited to small maps due to lack of the efficient management of large-scale map data applied with map matching. Thus,firstly the distributed Hidden Markov model is proposed. The model is able to calculate the Hidden Markov model which is a map matching algorithm with high accuracy on the distributed calculation framework by taking advantage of the severability of Viterbi algorithm. Without losing matching accuracy, the segmental management of map matching according to geographic location is realized,accelerating map matching 50% by cutting I/O cost.