Unsupervised Spatial Feature and Change Detection in RS Imaging
Abstract
Most of our work since the last interim report was spent on studying necessary adaptations of the software to various operational requirements, assisted by data runs to suggest their use, and studying and assessing recent statistical and neural theoretical developments for possible application. Main activities continued: (1) requirements analysis for development of prototype into a system for operational use; (2) re-designing the basic methodology implemented by the system, involving a clear separation of the main methodological elements; (3) differentiating project pursuit classification and pixel blending within classes; (4) conserving inspection mode to facilitate selecting (sub) classes and regions in a; (5) classified image for further analysis; and (6) continue study of spectral de-mixing, within (sub) classes, using RBF neural network technology.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1999
- Accession Number
- ADA372455
Entities
People
- R. J. Mokken
Organizations
- University of Amsterdam