The development of a spectral-spatial classifier for earth observational data

Abstract Over the last several years a classifier for earth observational image data has been under development which is intended to achieve improved performance by utilizing spatial characteristics of the data as an adjunct to multispectral ones. This paper provides an overview of the conception, development, evaluation and documentation of this spectral-spatial classifier. The research program leading to this classifier is described, the algorithms of the current implementation called ECHO are outlined, and results on its performance are summarized. These results show it to have improved accuracy, with greater computation efficiency, and only slightly increased operator complexity.