is quite
dreds of
herefore
facilities
involves
nce data
ind then
he form
ids. The
positive
general
features,
sis (e.g.
mple of
ad the
posed of
diverse
f all the
d at the
nts and
ee types
(b) non-
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002
linear mixture and (c) coatings.
50
0
Kaolinite Aline
a > us Alunite
e ©
S > =
7 400 —
e
S 359 - : >
9 ' 0
2 1 1 \ Q 180
S 300 À a8 E
o L| * -
o wd * 3
= wv 130
= 2.1600 * . ©
23080 *, | e
200.4 2.2100 M s 2.1560 2.3160
i Ï 7 T T i i
2 205.21 215.22 225 23 235 24 2.205 24 21322 225 26 231 24
Wavelength (in um) Wavelength (in um) ;
350 7
^. |Buddingtonite
S 300
©
x 250 -
s Figure 3. Comparison of single-pixel AVIRIS spectra
T ^ e 1. :
8 200 420240 Let (solid lines) and corresponding laboratory reflectance
8 i spectra (dashed lines) of kaolinite, alunite and
* e . . . . . . .
S una “2.1160, + buddingtonite from the Cuprite mining district; vertical
o . . .... .
X T bars indicate centre positions of absorption features
100 i— — | characterizing these minerals. (Van der Meer 1999)
2.2.08 21 215 22 225 2.3 235 24
Wavelength (in pm)
In a highly simplified form, the technique of spectral unmixing
uses a linear mixing model. The spectra of end members are
known (either from a spectral library or through the application
of statistical / procedures on the scene data) and observed mixed
pixel spectrum is known. From these data sets, the
abundance/relative proportion values of end members can be
obtained. Pixel-by pixel analysis is carried out to calculate the
amount (fraction) of each end member component in the
observed pixel. This results in fraction images, which show the
abundance and distribution of end member components in the
scene. These data provide information on the composition of the
surface and are also called ‘mineral maps’ (Fig. 4).
jarosite :
2.1.4 Limitations: The current method of mineral identification
and quantification using hyperspectral remote sensing data in
VIR-NIR-SWIR region have shown limitation owing to the
following main reasons:
1. On the Earth's surface, there is often a top surface cover of
vegetation, algae, lichen, soil, regolith, alluvium, moisture etc.,
which would mask the reflection characteristics of the bedrock.
Therefore, application of the technique is mostly limited to dry
desertic areas with bare outcrops.
2. Some of the minerals possess coatings, which would hinder
their identification.
3. Linear mixture modelling, assumed for unmixing the end-
members, may not be applicable in some cases; this would
Figure 4. Fraction maps of selected minerals (after Shang et al. 2002).
47]
