Performance of fuzzy RBFNN using FCM and FSC

This code uses a fuzzy extension of RBFNN using Fuzzy c means clustering and Fuzzy Supervised Classification.

Handling mixed pixel is a common problem for both multi as well as hyperspectral satellite data. This code proposes a fuzzy extension of Radial Basis Function Neural Network for classification of satellite data. The proposed method first estimates fuzzy membership values of satellite data using fuzzy-c-means algorithm. Similarly fuzzy supervisedclassification is performed on the same sattelite image using ground truth samples. Then from both FCM and FSC classified data sample selected for RBFNN. The procedure for connecting the FCM and FSC through RBFNN is explained in the title image.

Three experiments are performed with multi and hyperspectral satellite data, namely, Indian Remote Sensing Satellite-1A, LANDSAT-TM (Thematic Mapper) and Airborne Visible Infrared Imaging Spectrometer (AVIRIS) to compare the proposed FCM-RBFNN-FSC with basic FCM and FSC.