Section: Application Domains
Remote sensing is an image acquisition paradigm in which various pieces of information from the surface of the earth are obtained by airborne or satellite imaging. In the recent years the quality and variety of sensors available on the market for such remote acquisitions has drastically improved, leaving a large number of unsolved or difficult data processing tasks on the hand of data scientists. In particular, members of the PANAMA project-team have focused on processing hyperspectral images, that are images collected at a large number of wavelengths. The tasks that have been addressed are mainly spectral unmixing, that aims at identifying the various materials on the earth surface with their relative abundances, and denoising, that aims at removing structured noise from remote acquisitions that can be polluted by e.g. sensor movement or clouds. The spectra measured at each pixel of an hyperspectral image are heavily dependent on surface features (material, slop, material mixture type) which makes these data extremely rich and interesting.