Combining Attribute with Geometry for Automated Generalization and Schematisation

Citation

Leibovici DG, Swan J, Anand S & Jackson M (2008) Combining Attribute with Geometry for Automated Generalization and Schematisation. In: International Research Symposium on Computer-based Cartography. (Auto Carto), Shepherdstown, West Virginia, USA. AutoCarto 2008, the 17th International Research Symposium on Computer-based Cartography, Shepherdstown, WV, USA, 08.09.2008-11.09.2008. Cartography and Geographic Information Society. http://www.cartogis.org/docs/proceedings/2008/leibovici.pdf

Abstract
Map Generalization is the process by which coarse scale maps are to be derived from fine scale maps, balancing the amount of real-world information with visual confusion. This requires the use of operations such as simplification, selection, displacement and amalgamation of features that are performed subsequent to scale change. Map schematization can be regarded as a specific type of map generalization, focusing more on the readability, the purpose and the user context, including the visualisation device, for example the London Tube map. Advanced generalization functionality derived from the literature is now being found in commercial GIS software, but many research challenges remain. Recently, focusing on the attribute values of the geometrical objects, some research has been on thematic maps, such as demographic maps, soil maps, land cover and land use maps. For these situations, algorithms need to consider ontology associated with the theme and/or statistical clustering methods. What is happening to the attributes when performing a geometrical generalization or schematization of a map with polygons in which some attributes describe their semantics? Conversely what is happening to the geometries when, some kind of generalization based on attribute values and their spatial distribution, is performed? For instance, these questions concern sequential approaches, implying that one generalization either geometrical or attribute based, will force another generalization on the other characteristic. Investigating how to use the two main generalization steps in a more integrated approach will be the ultimate goal of this paper. Examples from land cover dataset and census dataset are used for this study.

Keywords
Automated generalization; map, schematization; classification; competing algorithm; entropy; census data; land cover