A Parallel Map-Matching Approach for Large Volume Floating Car Stream Data

Mp-matching floating car data is a fundamental task in traffic surveillance, traffic anomaly detection, and urban dynamic analysis. This study proposes a parallel map-matching approach to process streaming large volume floating car data. Considering the connectivity of a transportation network, the matching candidates are limited with a coarse spatial grid. A distance filter and a direction filter are combined to reduce the number of matching candidates. The trajectory between consecutive nodes is recovered with a shortest path list. The shortest path list in memory was developed to reduce the computation and speed up the matching process. A non-relational distributed database parallelizes the map-matching procedure. The performance of the presented approach was tested with large volume floating car data in Wuhan, China. It demonstrates that this method achieves 90.62% correct map-matching results. This efficiency could fulfill the needs of real-time traffic monitoring, and will benefit trajectory analysis.