19 June 2001

UNCERTAINTY ISSUES OF CONFLATION IN A DISTRIBUTED ENVIRONMENT

M. Cobb
University of Southern Mississippi
F. Petry
Tulane University
New Orleans, LA 70118
K. Shaw
Naval Research Laboratory
Stennis Space Center, MS 39529-5004

Abstract

We have previously considered issues relative to conflation in a geographic information system (GIS) based on the Vector Product Format (VPF) developed by the National Imagery and Mapping Agency (NIMA) as the Department of Defense relational database interchange standard for vector mapping. In this context we developed a knowledge-based system to provide input on both spatial and non-spatial properties of geographic features. Degrees of matching were generated for candidate features based on measurable, objective measures as well as subjective ones. Matches for non-spatial properties (attributes) were generated on the basis of similarity tables developed for the allowed VPF attribute sets. These tables were used in conjunction with a fuzzy combination function to provide the overall degree of matching of the candidate features' attribute/value sets. An expert system also generated weights for the combination function using rules that represented semantic interrelationships of feature attributes.

We are currently developing a distributed object-oriented spatial database at the Naval Research Laboratory. Within this context, conflation is a very evident concern, as the distributed environment entails possible issues of schema (metadata) merging, as well as specific feature data conflation. Our previous approach using an expert system for conflation is not directly feasible in a distributed environment. Hence, we have developed a model of conflation tailored to a distributed environment, for which uncertainty issues handled by our previous conflation system are fully accounted.

Sponsored by the Marine Corps Warfighting Laboratory.

Presented in GIS/LIS 98, November 9-12, 1998.
Naval Research Laboratory Contribution Number NRL/PP/7441 98-0010.
Conference Proceedings