Improving the accuracy of multi-scale building matching based on dynamic aggregation

Multi-scale building matching, which is crucial for spatial data updating, is complicated by the map generalization process. The reverse application of map generalization methods to improve multi-scale building matching has become a new research direction. This paper proposes a dynamic aggregation method. Firstly, large-scale buildings are combinatorially enumerated and dynamically aggregated into convex hulls, based on which the convex hull similarity to the small-scale building is calculated to ultimately select the combination with the highest similarity. Secondly, a Delaunay triangulation classification filtering method is employed to aggregate the candidate building combination and compute the overall similarity after aggregation. Finally, match success is determined by overall similarity, and the matching relationship is transferred to the original buildings involved in the aggregation. Experimental verification on two sets of datasets shows that after aggregation processing, the matching F1 indicators have increased by 28.57% and 9.28% respectively, reaching 94.62% and 90.44%.