Design of a modular feature extractor for hyperspectral images

Abstract

Hyperspectral image sensors capture surface reflectance across a range of wavelengths, which record detailed spectral information in terms of hundreds of bands. The classification of hyperspectral images has generated significant interest among researchers in the remote sensing community, as the large number of bands provides rich spectral information that can be used to classify objects, determine chemical components, or detect vegetation changes; all of which are useful in areas such as agriculture, geology, medicine, etc. However, due to dense sampling, some bands may contain redundant information; and sometimes, spectral information alone may not be su!cient to achieve the desired accuracy in results. E”orts have been made to describe all possible feature extraction mechanisms that add more information to multi-spectral images.
Therefore, in this work, we present the design of a tool that implements a uniform and configurable interface for the extraction of spectral and spatial features from hyperspectral images.