Automated Processing of Satellite Imagery Data:Test of a Spectral Classifier.

摘要

This report details the development and testing of algorithms for automated classification of Very High Resolution (VHR) cloud imagery from orbiting weather satellites. The imagery data are first processed into average wave number spectra. The spectral features are then input to a Bayesian quadratic discriminant classifier. The classifier is designed to work with VHR visual and IR data, but was tested here on VHR visual data alone. Comparison of the classifier to a blind processor using a priori probabilities shows accuracy of classification appreciably better than chance. (Author)