A belief function approach for using G.I.S. derived spatial context in satellite image understanding
One of the potentially important applications of geographic information systems is in providing spatial context information which can be useful for improving the classification of remotely-sensed satellite imagery. However, the integration of conventional image processing techniques with GIS and artificial intelligence approaches for image understanding is a difficult problem. A method for performing this integration is developed using expert system rules, triggered by geographic context for refining segment classification in land use mapping. The technique used is based on the Dempster-Shafer theory of evidential reasoning. Belief functions are computed by combining class likelihoods produced by image classifiers with additional class evidence derived from spatial context rules which are driven by a GIS. The basis of this technique is outlined and it is shown how image classification can be improved by deriving belief functions from specific expert system rules which use GIS spatial information as input.
Bibliographic Reference: Paper presented: European Conference on Geographical Information Systems, Amsterdam (NL), April 10-13, 1990
Availability: Available from (1) as Paper EN 35327 ORA
Record Number: 199011728 / Last updated on: 1994-12-02
Original language: en
Available languages: en