A comprehensive approach for the efficient acquisition and processing of hyperspectral images and sequences

Abstract

Despite the scientific and technological developments achieved during the last two decades in the hyperspectral field, some methodological, operational and conceptual issues have restricted the progress, promotion and popular dissemination of this technology. These shortcomings include the specialized knowledge required for the acquisition of hyperspectral images, the shortage of publicly accessible hyperspectral image repositories with reliable ground truth images or the lack of methodologies that allow for the adaptation of algorithms to particular user or application processing needs. The work presented here has the objective of contributing to the hyperspectral field with procedures for the automatic acquisition of hyperspectral scenes, including the hardware adaptation of our own imagers and the development of methods for the calibration and correction of the hyperspectral datacubes, the creation of a publicly available hyperspectral repository of well categorized and labeled images and the design and implementation of novel computational intelligence based processing techniques that solve typical issues related to the segmentation and denoising of hyperspectral images as well as sequences of them taking into account their temporal evolution.