Performance evaluation of different classifiers (Isoseg, Bhattacharyya, Maxver e Maxver-ICM), using CCD/CBERS-1 and ETM+/Landsat-7 fused images

Abstract

The aim of this study was to compare the performance of image classifiers (Isoseg, Bhattacharyya, Maxver and Maxver-ICM) based on an accuracy analysis (set percentage, area determination and Kappa coefficient), using as ground truth an edited thematic map. For this, pre-processing techniques (atmospheric, geometric and radiometric corrections), contrast enhancement (IHS data fusion and principal component analysis) and classification of CCD/CBERS-1 and ETM+/Landsat-7 images were done. Amongst all classifiers tested, Isoseg and Bhattacharyya presented best performance for the studied classes and the study area. It is anticipated that these results are relevant to environmental analyses based on orbital satellite data.