Improving Spatial-Spectral Classification of Hyperspectral Imagery by Using Extended Minimum Spanning Forest Algorithm

Abstract

Many researches have demonstrated that the spatial information can play an important role in the classification of hyperspectral imagery. Recently, an effective approach for spatial-spectral classification has been proposed using Minimum Spanning Forest (MSF) algorithm. Our goal is to improve this approach to the classification of hyperspectral images in urban areas. In the proposed method two spatial/texture features, using wavelet and Gabor filters, are first extracted. The Weighted Genetic (WG) algorithm is then used to obtain the subspace of hyperspectral data and texture features. They are then fed into a novel marker-based MSF classification algorithm. In this algorithm, the markers are extracted from the two spatial-spectral classification maps. To evaluate the efficiency of the proposed approach two image datasets, Pavia University acquired by ROSIS-03 and Berlin by HyMap, were used. Experimental results demonstrate that the proposed approach achieves approximately 17% and 14% better overall accuracy than the original MSF-based algorithm for these datasets, respectively.