VALUES FOR THE FUZZY -MEANS CLASSIFIER IN CHANGE DETECTION FOR REMOTE SENSING

We discussan approach to change detectionin digital remotelysensedimagery that relieson theFuzzyPostClassificationComparisontechnique. We usethefuzzy -means classifiertogetherwith theMahalanobisdistanceasthebasis for a metricof classmember ship for a individual pixel. We note that the valueof the fuzzyexponent in a fuzzy classifier is basedon the ratios of the reciprocals of the classMahalanobisDistances. The paper is an empirical investigationof this valueand concludeswith recommendationsfor thevalueof thefuzzyexponent .