Feature Extraction and Recognition of Two-Dimensional Data by the Method of Moments
Abstract
A pattern recognition system applicable to two dimensional data is presented, and training algorithms for generating pattern classifiers are surveyed. The method of moments is used by the system as a feature extractor. The Mahalanobis distance measure is presented as a criterion for the selection of moment pairs to be used as descriptors. Experiments conducted using simulated high resolution radar images demonstrate the effectiveness of the system using unstructured data. Classification results for the system are compared to those of human interpreters.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 15, 1977
- Accession Number
- ADA037445
Entities
People
- J. M. Harris
- R. C. Gonzalez
Organizations
- University of Tennessee