Adaptive procedures in multispectral pattern recognition

It is often necessary to have the statistical parameters periodically updated during the processing of remotely sensed multispectral scanner data. As is well known, one of the factors causing the degradation in classification accuracy is the variability of data due to the gradual changes in the spectral, spatial, temporal and other conditions. If an adaptation of these variabilities can be properly made, the performance accuracy of a processing system will undoubtedly be enhanced and the multispectral pattern recognition capabilities will be greatly extended to longer duration and/or larger area surveys.