Uncertainty Modeling

Sponsor: The University of Calgary
Project period: 1996
Principal Investigator: Dr. Ron Li
Project Team: A. A. Alesheikh

Description:

Uncertainties exist in all spatial databases. The main theme of this project was to devise appropriate models for and to propagate uncertainty through operations performed on the data. Uncertainty indicators can be modeled for six aspects. They are: lineage, positional uncertainty, attribute (thematic) uncertainty, logical consistencies, completeness, and temporal uncertainty. Positional and attribute uncertainty data are predominate in GIS databases, consequently requiring more attention to handle. The objectives are threefold. First, the spatial distribution of positional uncertainty for area entities (polygon) in GIS is modeled. An area object is a basic component for spatial analysis and is composed of several connected line segments (arcs). A model is devised for line segments. The combination of line segment uncertainties leads to the area object uncertainty estimator. Second, models are investigated for spatial distribution of thematic uncertainty for classified remote sensing data. Changes in natural resource data are gradual and fuzzy sets can accommodate the gradual transition, as well as the imprecise class definition. Lastly, The propagation of the above uncertainties through GIS processes will be investigated. Our research is also focused on tracking the above identified errors through overlay processes of GIS and represents the product uncertainty with the help of visualization programs.

E-mail:

<li.282@osu.edu>