Multiple Instance Bagging Based Ensemble Classification of Hyperspectral Images

Abstract

In this work, a novel approach is proposed for use in high dimensional spectral images by combining Multiple Instance (MI) Learning (MIL) with ensemble learning (EnLe). Ensemble learning models are constructed over random selections of instance and feature spaces by taken in to account of hyperspectral images' contextual information. Hyperspectral image with ground truth information is used for experimental results and comparative results are presented with State of art methods in MIL end EnLe.