Design And Implementation On Map Matching Algorithm Based On Big Data Platform

In order to meet the requirements of dealing with the accuracy and speed of massive GPS datas,this paper mainly completes the improvement of map matching algorithm and implementation of map matching algorithm in MapReduce parallel computing framework.Aiming at the first problem, the advantages and disadvantages of some existing map matching algorithms in real life are analyzed, and a map matching algorithm combining some traffic rules and the advantages of some map matching algorithms is proposed in this paper. The algorithm mainly uses the Hidden Markov Model (HMM) to model the map matching process, fully considering the topological relationship between the GPS information and the road network in the digital map,and improves the accuracy of the map matching algorithm to meet the needs for massive GPS datas mining. In this paper, the process of GPS data preprocessing including noise data removal, redundant data removal,missing data complement and drift data correction, the design and implementation of the improved map matching algorithm and a map matching algorithm based on topological information are finished.Compared with the precision and the processing speed of the map matching algorithm, it is concluded that the precision of the new algorithm is higher when dealing with the GPS data of the sampling frequency given in the article.At the same time, the parallelization of the map matching algorithm is studied as the second problem. The significance of this research is to improve the processing speed and save the time cost when dealing with a large number of spatiotemporal data. In this paper, the parallelization of GPS data preprocessing including parallelization of noise data removal,parallelization of redundant data removal, parallelization of missing data and correction of drift data, parallelization of the improved map matching algorithm, the single-threaded version of it and the multi-threaded version of it are designed and implemented. By comparing the correctness and time consumption of them, the design of the map matching algorithm based on the MapReduce calculation framework is obtained, and the time consumption of the parallel version is also shortened. As the data volume increases, this implementation has a greater advantage when compared with other two implements.