Online map matching by indexing approximate road segments

Given a typical real-world application, a large number of tracking points come at a high rate in a stream fashion and efficient online map matching is a pressing concern. In this paper, two map matching schemes are proposed, which make aggressive use of available main memory to achieve efficient matching. First, road network is partitioned into approximate segments, namely line or arc segments, by minimum description length (MDL) principle. Second, approximate segments are indexed into an optimized packed R tree, insuring that the minimal coverage and overlap are achieved. Based on this, we propose two online map matching schemes, namely Top Matching (TM) and kNN Refinement Matching (KRM), coupled with corresponding buffering strategies. Theory and experiment show that the times of accessing roads is significantly reduced to approximately a fraction of 3 to 6, demonstrating superior performance in terms of running time and matching quality.