Management of Uncertainty in Military Scene Analysis
Abstract
Research into modeling and managing uncertainty in military scene analysis is presented. A new method of logical inference was developed for the case where propositions are modeled by possibility distributions. This scheme was tested in a prototype fuzzy rule-based system for automatic target recognition. An information fusion technique based on the fuzzy integral was also developed. This was inserted into a numeric uncertainty propagation ATR prototype system. Fractal geometry was exploited for scene description and segmentation. Results concerning dimension calculation, texture description and segmentation, and surface orientation from fractal features is presented. Fast solutions to two problems in linear discriminant analysis are contained herein. Both techniques avoid the potentially disastrous errors from calculating large- cross product matrices. Finally, preliminary work on modifying confidence values for hypotheses based on external or scene derived context are presented. Keywords: Uncertainty; Fuzzy logic; Belief theory; Fuzzy integral; Rule-based automatic target recognition; Fractal geometry; Texture analysis; Linear discriminant analysis; Context.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 26, 1988
- Accession Number
- ADA219930
Entities
People
- James M. Keller
- Richard M. Crownover
- Robert W. Mclaren
Organizations
- University of Missouri