An Approach to Building Grouping Based on Hierarchical Constraints

Building generalization is one of the difficult operations in automated map generalization. It usually consists of two consecutive steps, i.e. dividing buildings into groups (also called building grouping) and performing generalization operations on building groups (also called generalization execution). This paper mainly focuses on the first step, which aims at proposing an approach to identify building groups for generalization. The proposed approach is based on the analysis of manual building grouping process, which leads to the basis of this approach that constraints are used hierarchically in building grouping process. In the approach, contextual features, such as road networks and river networks, and Gestalt factors, i.e. proximity, common orientation and similarity, are identified as constraints and they are used for building grouping in the same hierarchical way as manual grouping process: first, contextual features are used to partition buildings on the whole map into different regions; then, for each region, minimum spanning tree is employed to represent proximity relationships among buildings and three Gestalt factors are sequentially used as weights to segment a region into different groups. * Corresponding author. Tel.: +852 2766 5977; fax: +852 233