“…Recently, the multistage mode of hyperspectral data analysis has been playing a leading role, where performance of a classification system is supported by a set of analyses, such as segmentation, feature extraction and optimization, and dimensionality reduction. The empirical mode decomposition (EMD) for SVM classifier (Demir and Ertürk 2010), nonparametric feature extraction aided K‐NN classifier (Yang et al ), Bayesian learning‐based probabilistic sparse kernel aided (RVM, Mianji and Zhang ), posterior probability SVM (PPSVM, Wang et al ), kernel local Fisher discriminant analysis (KLFDA, Li W et al ), SVM with extended morphological attribute profiles and ICA (Kosaka et al ), ICDA framework (Villa et al ), adaptive MRF approach with SVM (Li S et al ), Canonical Bayesian classifiers (Zhang L et al ), mixture analysis and DT classifier (Delalieux et al ), and the feature selective linguistic classifier (Samadzadegan et al ) are becoming dominant approaches for hyperspectral data analysis. Imani et al () has proposed band clustering‐based feature extraction for the classification of hyperspctral data using limited training samples.…”