Road Intersection Detection Based on Direction Ratio Statistics Analysis.

Min Pu,Jiali Mao,Yuntao Du,Yibin Shen,Cheqing Jin

DOI: https://doi.org/10.1109/mdm.2019.00-46

2019-01-01

Abstract:Large collections of GPS trajectory data provide us unprecedented opportunity to detect the road intersection automatically. However, in the real-world scenarios, the precision of existing detection methods cannot be guaranteed due to severe challenges including (i) low-quality raw GPS trajectory data and (ii) the difficulty of differentiating intersections from nonintersections. To tackle above issues, we propose a novel twophase road intersection detection framework, called as RIDF, which is comprised of trajectory quality improving and intersection extracting. More importantly, through extracting candidate cells based on direction statistic analysis and refining the locations of intersections using hybrid clustering strategy, our approach can effectively detect road intersections of different size. An experimental evaluation on two real data sets extensively assesses the quality of RIDF method by comparing it with state-of-theart methods. Experimental results demonstrate that our proposal can overcome the limitations of existing methods and thus have better accuracy than the existing work.