A Novel Feature Matching Method for Matching OpenStreetMap Buildings with Those of Reference Dataset

Abstract

Numerous studies have attempted to assess the quality of OpenStreetMap's building data by comparing it to reference datasets. Map matching (feature matching) is a critical step in this method of quality assessment, involving the matching of polygons in the two datasets. Researchers commonly use two main polygon matching algorithms: 1) the buffer intersection method and 2) the centroid comparison method. While these methods are effective for the majority of OSM building footprints, they may not achieve high accuracy in complex situations. One possible reason is that both methods only consider the position of the OSM polygon compared to that of the reference polygon. To improve these matching algorithms and propose a more robust solution, this study proposes an algorithm that considers shape similarity (using average distance method) in addition to position similarity to better identify corresponding polygons in the two datasets. The experiment results for five cities in the Province of Quebec indicate that the proposed algorithm can reduce the matching error of previous map matching algorithms from approximately 8% to approximately 3%. Furthermore, the study found that the proposed polygon matching algorithm performs more accurately than previous methods when buildings consist of multiple polygons.