Boosted Band Ratio Feature Selection for Hyperspectral Image Classification

Abstract

Band ratios have many useful applications in hyperspectral image analysis. While optimal ratios have been chosen empirically in previous research, we propose a principled algorithm for the automatic selection of ratios directly from data. First, a robust method is used to estimate the Kullback-Leibler divergence (KLD) between different sample distributions and evaluate the optimality of individual ratio features. Then, the boosting framework is adopted to select multiple ratio features iteratively. Multiclass classification is handled by using a pairwise classification framework. The algorithm can also be applied to the selection of discriminant bands. Experimental results on both simple material identification and complex land cover classification demonstrate the potential of this ratio selection algorithm