Exploring support vector machine in spectral unmixing

Spectral unmixing is an important technique of hyperspectral imagery processing. The traditional iterative processing of least squares linear spectral mixture analysis is of heavy computational burden. In this paper, a simple distance measure is proposed based on support vector machine (SVM). The method is free of iteration and dimensionality reduction, with very low complexity. In addition, SVM model is applied to accommodate the variations within the relative pure pixels by using multiple pure samples instead of single endmember for one class. Experimental results show the effectiveness of the proposed methods.