AUTOMATED MATCHING CROWDSOURCING ROAD NETWORKS USING PROBABILISTIC RELAXATION

Abstract

Geospatial data matching is an important prerequisite for data integration, change detection, and data updating. Presently, crowdsourcing geospatial data is drawing great attention with its significant potential for geospatial data updating and Location Based Services, etc. To explore the availabilities of crowdsourcing geospatial data, the paper proposes a heuristic probabilistic relaxation road matching method, named PRRM. It starts with an initial probabilistic matrix according to geometric dissimilarities and then integrates the relative compatibility coefficient of neighbouring candidate pairs to update the previous matrix. Finally, the initial 1:1 matching pairs are selected based on probabilities calculated and refined based on the structure similarity of the selected matching pairs, then a matching growing process is implemented to find M: N matching pairs. Two experiments between OpenStreetMap and professional data show that our method achieves good performance in matching crowdsourcing and professional data with non-rigid deviations and inconsistent structures. Moreover, the proposed method is independent on matching direction and could handle 1: 0 ( Null), 1: M and M: N matching.