A New Approach to Compute Vector Based Morphological Features for Classification of Hyperspectral Image

Analyzing hyperspectral images using supervised classification is challenging, primarily because of the higher spectral dimensions and rich, complex spatial information. This article proposes a vector strategy to compute multivariate morphological profiles for feature extraction to classify hyperspectral images using support vector machines (SVMs). The proposed vector ordering strategy is based on Dilation-distance to compute multivariate morphological operations. The features obtained from computing a multi-channel Morphological profile using the proposed vector ordering strategy are used for the classification of the data. Experimental results utilizing hyperspectral data “Pavia University” show that classification using the multichannel profiles can surpass multi-channel profiles using other vector ordering strategies when extracting pivotal features for classification.