Formalization and data enrichment for automated evaluation of building pattern preservation

Automated evaluation of generalization output relies largely on well defined map specifications or cartographic constraints to formalize quality criteria or user requirements, and on enriched information to fully automate the process. Previous studies suggest that the formalization and evaluation of legibility constraints (i.e. those improving the readability of maps) are relatively easier than preservation constraints (i.e. those preserving important geographic characteristics and patterns). Patterns are important structures and should be taken into account by the evaluation process. This paper aims at a methodology in which the preservation of building patterns can be evaluated automatically. Three major difficulties in the process are identified and addressed: (1) pattern classification, characterization and detection, (2) pattern matching, and (3) constraint formalization. In addition, the knowledge of transition events describing allowed changes to building patterns at scale transitions is obtained via studying existing topographic maps (from 1:10 k to 1:100 k). Based on this knowledge, automated matching of corresponding building patterns is done much easier. Finally, the proposed methodology is implemented and validated by being applied to the evaluation of an interactively generalized dataset against its initial dataset. The results show the potential of the methodology and that the generalized patterns tend to remain the same or become more regular with respect to the initial ones. We also identify further improvement for practical use in an overall evaluation process to indicate acceptable generalization solutions.