Automatic Recognition of Building Cluster Patterns for Spatial Scale Transfer

Building patterns are salient urban and rural structures for map readers (e.g. urban geographers and tourists) to understand the organization, function, history and development of a city or city parts (Lévi-Strauss, 1963). To represent and preserve the patterns at various scales for different applications has always been a challenge for cartographers, as it requires knowledge and experience on the implicit patterns and also skills to generalize them. It is even harder for computers to automate the generalization of city maps where building patterns are one of the major structures, because the patterns are not explicitly stored in geospatial datasets. Therefore, automatic recognition of building patterns from geospatial data is required to better support the automated generalization of city structures.