ADAPTIVE STATISTICAL PATTERN CLASSIFIERS FOR REMOTELY SENSED DATA

摘要

A new technique for the adaptive estimation of non-stationary statistics necessary for Bayesian classification 1s developed. The basic approach to the adaptive estimation procedure consists of two steps: (1) an optimal stochastic approximation of the parameters of interest and (2) a projection of the parameters 1n time or position. A divergence criterion 1s developed to monitor algorithm performance. Comparative results of adaptive and non-adaptive classifier tests are presented for simulated four dimensional spectral scan data.