Abstract
A new technique to classify remotely sensed data was developed on the basis of the fuzzy theory. The classification was conducted to the cells which consist of some hundreds of pixels.
A membership function was derived from the histogram, i.e. frequencies of the occurrence of each of the digital numbers (DN) within a cell, since the hitogram is a probability function itself with regard to each of DN. Similarity, distance and likelihood between cells were defined using the membership function. A supervised classification was carried out by applying this tequnique to LANDSAT TM data. An unsupervised classification was also performed using this technique with a merging process of cells to obtain clusters. The classification results were displayed taking the 1st and the 2nd largest likelihoods into account.
