New index to remove from a spectral mixture an unknown spectral component

• Published in: Acta astronautica Vol. 56; no. 4; pp. 471 - 475
• Main Authors: Yanev, T.K., Choudhury, B.J., Mishev, D.N.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-02-2005
• ISSN: 0094-5765; 1879-2030
• DOI: 10.1016/j.actaastro.2004.05.071

More than one spectral classes are usually found within a pixel when measuring the spectral reflectance (the spectral reflectance coefficients (SRC) in particular) of natural formations in the visible (VIS) and near infrared (NIR) spectral ranges. The spectral mixture “soil-vegetation” is a typical example. The decomposition of this mixture is strongly impeded because only the mixture SRC are usually available, the soil SRC and the vegetation coverage coefficient
p not being known in advance. This is especially true for remotely sensed data in field experiments. Under the assumption that the soil SRC are a linear function of the wavelength two SRC transforms were defined and studied in previous papers of the authors. The second transform is the
θ coefficient, which is free of the unknown linear soil component as well as of the unknown coverage coefficient
p. In previous studies of the authors
θ was examined in the case when the soil SRC may be approximated by one straight line. In this work,
θ was obtained and studied for soil types that may be satisfactory piece-wise approximated by two different crossing straight lines within the VIS and NIR ranges. Thus, the
θ coefficient may be employed to eliminate the unknown soil SRC from the measured mixture SRC in the case when the soil SRC are nonlinear wavelength function which may be fitted by two straight lines. Examples of USA and Bulgarian soil types are given in the study, which meet this requirement.