OBTAINING BUILDING GROUPS IN URBAN BLOCKS BY EMPLOYING DIFFERENT CLUSTERING APPROACHES

Abstract

Spatial and cartographic modelling and analysis studies such as generalisation, thematic mapping and multi-scale representation usually require extracting relevant groupings of buildings. For this purpose, minimum spanning tree (MST), Delaunay subgraph, DBSCAN and Chameleon clustering algorithms are employed in order to detect local building clusters in urban areas. Effectiveness of these approaches are tested using a topographic dataset at 1:25 000 scale. First, urban blocks are created to constrain clustering process from topological aspect. Second, a proximity matrix is computed with the minimum Euclidian distances between building polygons. Finally building groups are obtained from four methods using the proximity matrix: MST Adaptive Spatial Clustering based on Delaunay Triangulation (ASCDT), Chameleon and DBSCAN algorithms. Findings of the study demonstrate ASCDT and DBSCAN are superior in finding meaningful building groups in urban blocks.