Unsupervised classification for remotely sensed data using fuzzy set theory

Fuzzy interpretations of data structures are a very natural and intuitively plausible way to formulate and solve various problems such as uncertainty, vagueness, decision making etc. The concept of fuzzy set theory without a priori assumption is used to devise a novel algorithm to carry out fuzzy symbolic classification of remotely sensed data (IRS 1B Satellite). The proposed algorithm involves two stages. In the first stage, the authors convert the data in to symbolic form, which involves data reduction followed by a new concept of finding the number of classes in the data based on the farthest neighbor index. In the second stage, fuzzy descriptions on symbolic objects of remotely sensed data is developed using membership function. Membership function is calculated using seed points determined from the farthest neighborhood concept instead of usual fuzzy means. Further classification is done, using fuzzy membership value. The classification results of IRS 1B satellite data covering Hyderabad City is encouraging. Results signify that fuzzy classification is more logic and more powerful than hard classification