Vector road network map play an important role in both intelligent transportation systems and navigation systems. However, many of the vector map data are not completely open to the outside world. And traditional vector road maps are obtained by surveying and mapping personnel using professional instruments in the field environment. In this way, it has high cost, long cycle, and may not be able to fully keep up with the current urban changes. In addition, sometimes the taxi GPS trajectory and the map data given are poorly mapped, making it difficult to match the trajectory to the map of the road segment. In this paper, we propose a series of methods to dig out massive taxi trajectory and build a system can generate vector road network map automatically. In our approach, a grid method is used to divide the two-dimensional space into grids of uniform size, and each grid records the trajectory information that passes through it. Then, a Top-N similarity algorithm is devised to find similar tracks to form driving paths in grid. Last we designed a Path-Merge algorithm to merge overlapping driving paths to form vector roads and intersections, besides, we adopt Geohash algorithm to accelerate retrieval speed. We build the system based on real world taxi trajectory dataset collecting over 500 thousand trips one day in period of one week. The result of vector road data is complete and accurate compared to real vector road data, which demonstrate our methods are effective and outperform.

CICTP