REMOTE SENSING TECHNOLOGY

Fully constrained linear spectral unmixing algorithm for hyperspectral image analys

A. Yu. Denisova^ab, V. V. Myasnikov<sup>ab

a</sup> Samara State Aerospace University
^b Image Processing Systems Institute, Russian Academy of Sciences

Abstract: In this article, a novel linear spectral unmixing algorithm is proposed and analyzed. The linear spectral mixture defines a model of pixels for hyperspectral images by means of spectral signatures. A set of spectral signatures is assumed to be known. Constraints are imposed on the spectral mixture coefficients: the sum of the coefficients is equal to unity and each coefficient is nonnegative. The results of the algorithm quality and speed analysis are described in the paper.

Keywords: hyperspectral images, linear spectral mixing, constraints, hyperspectral analysis, least squares method.

Received: 06.11.2014