AN IMPROVED SPECTRAL–SPATIAL CLASSIFICATION APPROACH FOR HYPERSPECTRAL DATA

Recently, an effective approach to spectral-spatial classification of hyperspectral images has been proposed based on Minimum Spanning Forest (MSF) grown from automatically selected markers using Support Vector Machines (SVM) classification. This paper aims at improving this approach by using the optimal trained samples into marker selection process. In this study, the markers are extracted from the classification maps obtained by SVM optimized and are then used to build the MSF. The optimization algorithms used to select optimal trained samples, are the Principal Component Analysis (PCA), Minimum Noise Fraction (MNF) and genetic algorithm (GA). To evaluate the proposed approach, three benchmark hyperspectral data sets, the Pavia University dataset, the Telops dataset and the Berlin dataset are tested. Experimental results show that the proposed approach using GA achieves an approximately 4%, 3% and 2% overall accuracy higher than the original MSF approach for the Pavia University, Telops, and Berlin datasets, respectively.