People usually communicate among each other about geographic space in a casual way, using imprecise spatial concepts and terms. When interacting with geospatial databases, however, they are forced to adopt much stricter spatial concepts that are implemented through discrete data structures and Boolean operators. We propose to develop measures for spatial similarity that overcome the shortcomings of traditional methods. Our similarity measures are based on spatial relations and attributes. Spatial relations will be used to capture the distribution of spatial objects through a multi-resolution model, allowing for coarse as well as more detailed analysis of topological, directional, and metri cal relations, while attribute similarity will be measured through a semantic network of feature classes. Through an integrated model of these similarity measures we will be able to capture the neighborhood of geospatial objects, giving us a better-quality model for describing change than either of the models alone. The formalization of the simila rity measures will be complemented by software prototype developments.
Models for Assessing Geometric and Semantic Similarity
Putting Similarity Assessment into Context
Matching Functions with the User's Intended Operations
A Combined Approach to Semantic Similarity Across Multiple Ontologies
Similarity Assessment between Cardinal Directions
Similarity Assessment in Spatial- Query-by-Sketch
Last updated on October 29, 1999.
[ Geographic Databases | Spatial Database Research Group | NCGIA Maine ]