A NEW SPECTRAL-SPATIAL FRAMEWORK FOR CLASSIFICATION OF HYPERSPECTRAL DATA

In this paper, an innovative framework, based on both spectral and spatial information, is proposed. The objective is to improve the classification of hyperspectral images for high resolution land cover mapping. The spatial information is obtained by a marker-based Minimum Spanning Forest (MSF) algorithm. A pixel-based SVM algorithm is first used to classify the image. Then, the marker-based MSF spectral-spatial algorithm is applied to improve the accuracy for classes with low accuracy. The marker-based MSF algorithm is used as a binary classifier. These two classes are the low accuracy class and the remaining classes. Finally, the SVM algorithm is trained for classes with acceptable accuracy. To evaluate the proposed approach, the Berlin hyperspectral dataset is tested. Experimental results demonstrate the superiority of the proposed method compared to the original MSF-based approach. It achieves approximately 5 % higher rates in kappa coefficients of agreement, in comparison to the original MSF-based method.