Symbolic Model-Based SAR Feature Analysis and Change Detection
Abstract
The feasibility a model-based approach to feature classification and change detection in synthetic aperture radar imagery is investigated. The models include quantitative, qualitative and relational aspects of feature signatures. The intended Phase HI classification system will consist of an automated model-driven linked to an image understanding library such that during mated classification, the model-based processor will call various library routines as required to reach solutions. Phase I effort consisted of: (1) Investigation of various methodologies for rule-oriented model-based approaches; (2) Development and test results of a numerical scoring function for tracking uncertainties during rule-based processing; (3) Developmental and specification of the required image processing library functions; (4) Development and testing of a 120 rule prototype classification system; and (5) Results from testing the combination rule prototype and image processing library on four SAR images. The initial testing of the 120 rule prototype over the SAR data sets resulted in a 71% correct classification rate overall.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1992
- Accession Number
- ADA251277
Entities
People
- James C. Curlander
- Wolfgang Kober