Introduction
Statistical classification of hyperspectral data is challenging because the input space is high in dimension and correlated, but labeled information to characterize the class distribution is typically limited. The resulting classifiers are oftern unstable and have poor generalization.
In this ensemble approach, we focus on creating an ensemble of diverse classifiers by using random sampling on the sample set (bagging), input space (random subspace method) and both of these two spaces (random forests). This ensemble method uses either simple or weighted voting scheme to determine the outcome. The goal of this approach is to achieve not only high classification accuracy but also good generalization.