Authors

J. M. Haut, M. E. Paoletti, A. Paz-Gallardo, J. Plaza, A. Plaza

Proceedings of the 17th International Conference on Computational and Mathematical Methods in Science and Engineering, CMMSE

Volume 3. Págs 1063-2321

17th International Conference on Computational and Mathematical Methods in Science and Engineering (CMMSE'17), July 4-8, 2017

ISSN: 2312-0177

ISBN: 978-84-617-8694-7

Abstract

Classification of remotely sensed hyperspectral images is a challenging task due the enormous  amount  of  information  comprised  in  these  images,  that  contain  hundreds of  continuous  spectral  bands.    This  creates  a  need  to  develop  new  techniques  for hyperspectral classification using high performance computing architectures.  Despite the  availability  of  multiple  algorithms  adapted  to  parallel  environments  (such  as multicore  computers  or  accelerators  like  field  programmable  gate  arrays  or  graphics processing  units,  the  application  of  cloud  computing  techniques  has  not  been  as widespread, although there are many potential advantages in exploiting cloud computing architectures  for  distributed  hyperspectral  image  analysis.   In  this  paper,  we  present a  cloud  implementation  (developed  using  Apache  Spark)  of  a  successful  technique for hyperspectral image classification:  the multinomial logistic regression probabilistic classifier.  Our experimental results suggest that cloud computing architectures allow for the efficient classification of large hyperspectral image data sets.