A Comparison of the Performance of Fuzzy Algorithm versus Statistical Algorithm Based Sub-pixel Classifier for Remote Sensing Data

It is found that sub-pixel classifiers for classification of multi-spectral remote sensing data yield a higher accuracy. With this objective, a study has been carried out, where fuzzy set theory based sub-pixel classifiers have been compared with statistical based sub-pixel classifier for classification of multi-spectral remote sensing data.Although, a number of Fuzzy set theory based classifiers may be adopted, but in this study only two classifiers are used like; Fuzzy c-Means (FCM) Clustering, Possibilistic c-Means (PCM) Clustering. FCM is an iterative clustering method that may be employed to partition pixels of remote sensing images into different class membership values. PCM clustering is similar to FCM but it does not have probabilistic constraint of FCM. Therefore, the formulation of PCM is based on modified FCM objective function whereby an additional term called as regularizing term is also included. FCM and PCM are essentially unsupervised classifiers, but in this study these classifiers are applied in supervised modes. Maximum Likelihood Classifier (MLC) as well as Possibilistic Maximum Likelihood Classifier (PMLC), the new proposed algorithm have been studied as statistical based classifier. All the algorithms in this work like; FCM, PCM, MLC and PMLC have been evaluated in sub-pixel classification mode and accuracy assessment has been done using Fuzzy Error Matrix (FERM) (Binaghi et al., 1999). It was observed that sub-pixel classification accuracy various with different weighted norms.