Geoscience Reference
In-Depth Information
end-members (Tompkins et al.
1997
). In order to use a linear mixture model, the
spectral reflectance of the “pure” end-members should be measured. Ideally,
ground-based reflectance spectra would be acquired from the known minerals and
rocks to produce accurate end-members. Since we are concentrating the lunar
surface, we are not able to provide the specific high-precision ground truth informa-
tion/spectra at this present situation. Spectral mixture analysis (SMA), a technique
based on modeling image spectra as the linear combination of end-members, has
been used to derive the fractional contribution of end-member materials to image
spectra in a wide variety of applications. An end-member is a “pure” spectrum of
a material or target and has a unique spectral signature. Image end-members are
pure pixels from image itself. The end-member selection included the following
components: plagioclase, ortho-/clinopyroxene, and olivine in the mineral class and
low-, medium-, and high-Ti basalts in rock classes. Performing spectral unmixing
requires appropriate end-member selection, since unsuitable end-member may lead
to meaningless fraction mapping. In this paper, seven end-members have been
identified and extracted from the image data for spectral unmixing analysis.
The bidirectional reflectance spectra of low-, medium-, and high-Ti basaltic soils
acquired from various grain sizes such as <10, 10-20, 20-45, and <45 m. Based
on the grain size variation, there is not much variation observed in the spectral
curves apart from the broad weak absorption near 1,000 nm for the grain size of
20-45 m sample. To establish the basaltic end-member's spectra, the RELAB
spectra of low-, medium-, and high-Ti basalts of various grain sizes have been
taken and averaged. The spectral values are averaged for the various grain sizes, and
one spectral end-member is derived for each basaltic soil. The average of the bulk
chemical composition in oxides and in modal abundance of minerals in the various
grain sizes of low Ti (15071-52), medium Ti (12030-14), and high Ti (71501-35) is
given in Tables
2.1
and
2.2
, respectively. The spectral profiles of RELAB mineral
end-member (a) and basaltic end-members (b) are shown in Fig.
2.5
. The spectra
from 450 to 2,500 nm are used as end-member to estimate the abundance of the
various basalts in the Orientale basin basaltic terrain. Apart from the basalts, the
plagioclase, ortho-/clinopyroxene, and olivine spectra from RELAB are also utilized
in the present study.
2.3.1.5
Linear Spectral Unmixing
Spectral unmixing methods have been proven useful in the application of imaging
spectroscopy for geological studies. It is understand that, grain size effects intimate
mixtures and mineral coatings, so other spectroscopic spectral unmixing methods
have proven useful in the application of imaging spectroscopy to geological studies.
Linear spectral unmixing (LSU) has been proposed for the analysis of hyperspectral
images to compute the fractional contribution of the detected end-members to each
pixel in the image. Linear spectral unmixing is used to discriminate one type of
material from another. It provides an estimation of how much material available
in a pixel and a method of classifying broad categories of materials in a pixel.
Search WWH ::
Custom Search