Segmented and non-segmented stacked denoising autoencoder for hyperspectral band reduction

摘要

In order to make hyperspectral image classification compu- tationallytractable, it is often necessary to select the most informative bands insteadto process the whole data without losing the geometrical representation oforiginal data. To cope with said issue, an improved un- supervised non-lineardeep auto encoder (UDAE) based band selection method is proposed. The proposedUDAE is able to select the most infor- mative bands in such a way that preservethe key information but in the lower dimensions, where the hiddenrepresentation is a non-linear trans- formation that maps the original space toa space of lower dimensions. This work emphasizes to analyze what type ofinformation is needed to preserve the hierarchical UDAE representation whileselecting a sub- set from original space. Our experiments on publicallyavailable hyper- spectral dataset demonstrate the effectiveness of UDAE method,which equates favorably with other state-of-the-art methods.