In this study the fuzzy filtering algorithm enunciated by the authors is employed as a classifier to accurately label spectral classes distinguished within the imagery according to their corresponding land-cover condition. The study area is composed of two parts, a training site and a test site, each divided into small square meshes of the size 200m×200m for which the proportion of land-cover category is required to be estimated. Prior to land-cover classification category specific brightness spectra dealt with as fuzzy numbers are identified, with centers and widths separately determined by the Kalman filtering technique. Classification results show that the proposed method outperforms the simple Kalman filtering method as well as two conventional methods or maximum likelihood method and linear discriminant function method in five category classification case. As regards thirteen category case, the proposed method yields estimates with satisfied accuracy comparable to or better than the results brought about by the Kalman filtering method.
