04
1er
2m
urs
led
18
lue
ed
ain
on
18
eas
nm
rth
nes
rea
| 6)
in
| in
-H
her
the
ion
ese
ind
ica
/ in
lost
or
aps
t of
this
less
ong
Mg-
a of
] to
/IR.
nm
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
belong to group of AIOH bearing minerals e.g. Pyrophyllite and
Alunite. Jarosite represents sulphite minerals and has absorption
feature at 2260 nm. Minerals with absorption feature near
2327.5 nm belongs either to CaCO; or MgOH bearing minerals
— this group of minerals represent Calcite, Chlorite or Talc.
Group of minerals which has absorption feature near 2205 nm is
represented by Montmorillonite, Kaolinite, Muscovite or Illite.
Whilst will be not possible to map particular minerals using
ASTER imagery, it will be possible to differentiate between
mentioned mineral groups.
Selection of endmembers. Selecting of different types of
endmembers which represent spectrally most extreme pixels in
the scene has been provided by support of pixel purity index
technique (PPI). PPI is a means of finding the most “spectrally
pure” pixels typically correspond to mixing endmembers. The
PPI is computed by repeatedly projecting n-dimensional
scatterplots onto a random unit vector. The extreme pixels in
cach projection are recorded and the total number of times each
pixel is marked as extreme is noted (ENVI User’s Guide).
The spectra with specific spectral characteristics or of areas
important for geological setting has been also studied, but have
not been noted as extreme and are not used for spectral
analyses. The image scene has been divided into several areas
according to conclusions of image interpretation and has been
focused to map geological and geographical phenomena
occurring on peninsula.
Matched filtering is an ENVI software technique which
performs partial unmixing — finding the abundances of user
defined endmembers. Not all of the endmembers in the image
need to be known. This technique maximizes the response of
the known endmember und suppresses the response of the
composite unknown backgroung, thus “matching” the known
signature. It provides a rapid means detecting specific materials
based on matches to library or image endmember spectra and
does not require knowledge of all endmembers within an image
scene (ENVI User’s Guide). Bands 4-9 have been processed by
Matched Filtering in order to map area according to its SWIR
spectral signatures. Following spectra have been collected:
a/ WpoolW — which represents the pixels with 2205 nm
absorption, no pixels with absorption at 2205 and without
absorption at 2237.5 nm at the same time have been found in
the image scene.
b/ Solonchak — represents pixel with deepest absorption feature
at 2165 nm
c/ Dunes bright — represents pixels with absorption at 2165 nm
and with no absorption at 2237.5 nm
d/ Top of duine — represents pixels with absorption at 2165 nm
and 2237.5 nm
e/ NE pool — represents pixels with deep absorption at 2237.5
nm
Spectral Library Viewer
TUN
= Solorngoehgk
= ]Duines prigkd
& Hop d. duinie
2 I nl
= Sven
= Ed
% TM
m
€ ] [or
o
"B co 20
Yravelenglh
Figure 7. Spectra of selected endmembers in the SWIR.
The study area has been successfully mapped using matched
filtering procedure. Results of matched filtering are shown as
three endmembers colour composite (in red, green respectively
blue colour) where higher matches are represented by brighter
pixels. Figure 8 shows western part of Red Series outcrops
where pixels with different spectral signatures are mapped. The
randomly collected spectra of bright pixels demonstrate that
technique is able to classify image according to spectra
signatures (plot).
Wavrlenalh
Figure 8. The matched filtered image of west part of Red Series
shows different spectral signatures along the profile (red line).
Red pixels (both in image and plot) have absorption feature at
2205 nm and 2237.5 nm, green and blue have absorption both
2165 nm and 2237.5 nm, when green one have deeper the first
absorption feature and blue ones the later.
3. CONCLUSIONS
Analysis of ASTER spectral reflectance data of Cheleken
Peninsula provides promising results. ASTER imagery allowed
us to carry out preliminary geological investigation of the area
and find encouraging research methods in terms of mapping the
main geological phenomena occurring around the peninsula.
False-color composite images are useful to get an overview
about the area and map different types of rocks and land covers.
Band rationing is a mean to map Red Series (ASTER band 1/
band 2) and distribution of hydroxyl and carbonates bearing
minerals. Thanks to Matched Filtering technique surprising
facts have been discovered — two types of areas influenced by
649
