Share Email Print

Proceedings Paper

A novel hardware-friendly algorithm for hyperspectral linear unmixing
Author(s): Raúl Guerra; Lucana Santos; Sebastián López; Roberto Sarmiento
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Linear unmixing of hyperspectral images has rapidly become one of the most widely utilized tools for analyzing the content of hyperspectral images captured by state-of-the-art remote hyperspectral sensors. The aforementioned unmixing process consists of the following three sequential steps: dimensionality estimation, endmember extraction and abundances computation. Within this procedure, the first two steps are by far the most demanding from a computational point of view, since they involve a large amount of matrix operations. Moreover, the complex nature of these operations seriously difficult the hardware implementation of these two unmixing steps, leading to non-optimized implementations which are not able to satisfy the strict delay requirements imposed by those applications under real-time or near real-time requirements. This paper uncovers a new algorithm which is capable of estimating the number of endmembers and extracting them from a given hyperspectral image with at least the same accuracy than state-of-the-art approaches while demanding a much lower computational effort, with independence of the characteristics of the image under analysis. In particular, the proposed algorithm is based on the concept of orthogonal projections and allows performing the estimation of the number of end- members and their extraction simultaneously, using simple operations, which can be also easily parallelized. In this sense, it is worth to mention that our algorithm does not perform complex matrix operations, such as the inverse of a matrix or the extraction of eigenvalues and eigenvectors, which makes easier its ulterior hardware. The experimental results obtained with synthetic and real hyperspectral images demonstrate that the accuracy obtained with the proposed algorithm when estimating the number of endmembers and extracting them is similar or better than the one provided by well-known state-of-the-art algorithms, while the complexity of the overall process is significantly reduced.

Paper Details

Date Published: 20 October 2015
PDF: 15 pages
Proc. SPIE 9646, High-Performance Computing in Remote Sensing V, 96460I (20 October 2015); doi: 10.1117/12.2196109
Show Author Affiliations
Raúl Guerra, Univ. de Las Palmas de Gran Canaria (Spain)
Lucana Santos, Univ. de Las Palmas de Gran Canaria (Spain)
Sebastián López, Univ. de Las Palmas de Gran Canaria (Spain)
Roberto Sarmiento, Univ. de Las Palmas de Gran Canaria (Spain)

Published in SPIE Proceedings Vol. 9646:
High-Performance Computing in Remote Sensing V
Bormin Huang D.D.S.; Sebastián López; Zhensen Wu; Jose M. Nascimento; Boris A. Alpatov; Jordi Portell de Mora, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?