Effect Of Feature Extraction And Classification Method On Hyperspectral Image Classification Accuracy

In this paper, the effect of different hyperspectral images feature extraction techniques in ANN and SVM classifiers is investigated. While a high accuracy and efficiency for HLFE method was shown in ANN classifier in the literature, in this study it is shown that using the RBF kernel in SVM provides increased accuracy for PCA and KPCA meanwhile poor classification accuracy is achieved for HLFE. Therefore it is shown in this paper that it is important to take feature extraction and classification methods together into account, as one feature extraction method might outperform the other for a certain classification approach but might provide poor performance for another. This is an important point, usually neglected in the literature.