Classification of hyperspectral data using grey model

The main objective of this paper is to perform feature extraction over Hyperspectral data by means of the grey model. The wavelet transform was exploited as a tool to transform the spectral information into time-frequency domain. The grey model coefficients were generated from the wavelet coefficients set. The extracted grey model parameters were fed to the Support Vector Machines (SVM), and also to the Kth- Nearest Neighbor, Decision Tree, Discriminant Analysis, and finally Naïve Bayes classifiers. The obtained results over the benchmark AVIRIS dataset were compared. They show that the achieved performance using the developed method is either superior or comparable to that of the spectral signature alone depending on the classifier and its robustness against the curse of dimensionality.