On spatial information retrieval and database generalization

Spatial data are often stored in a database with the finest level of details. The problem of cartographic generalization of spatial data concerns deriving spatial data with proper level of details when the data are used for an application. The generalization process is not simply a process of selecting objects or parts of objects suitable for an application. It should maintain the semantic information carried with the spatial data, such as topological relationships among objects and the overall "look and feel" of the data after generalization. Most difficulties of map generalization come from the fact that the importance of objects, which determines whether an object should be included and how it should be visualized in the generalized data set, is decided not only by their sizes, but also the "meaning" of these objects to the application. The latter is application-dependent, and often subjective. As a classic problem in cartography, maps at different scales are pre-derived for different applications, with intensive human involvement. The performance of spatial data generalization is a secondary issue to the quality of the maps generalized, With more and more computerized applications making use of shared spatial data sources, and the increasing number of applications emerging from Internet-based spatial applications (such as map-based navigation), efficient spatial data generalization becomes a critical issue. At the same time, it becomes less and less likely to have human intervention or have pre-materialized spatial data generalization for all applications in the Internet environment. In this paper, we investigate issues in efficient spatial data generalization, with a unique focus on spatial database design and spatial operations processing. We propose several types of extensions and novel applications of z-values, a commonly used spatial indexing mechanism, to achieve our goals.