scene. Block
best lithologic
1, followed by
) PCs appear
This gave
mposite with
ich as basal
amorphosed
etc. are wel
nts such a
differentiated
nglomerates.
ell enhanced
raw data orf
Graphite rich
parable from
quence. The
— MÀ
1d morphologé
obtained enhancement by using correlation coefficient
instead of band covariance matrix is also supported by the
improvement in signal-to-noise ratio. Following Ready &
Wintz (1973) signal-to-noise ratio (SNR) may be determined
as the ratio between the first eigen value and the maximum
spectral band covariance. For the Ariab sub-scene;
using covariance matrix (table 1);
SNR = 1562.29/692.95 = 3.53 dB
whereas using correlation coefficient (table 2);
SNR = 5.65/1.0 = 7.52 dB
It can be seen that the obtained improvement is in the order
of 3.99 dB, signifying the superiority of the standardized PC
transformation. Furthermore, it should be noted that
topographic expressions are well preserved in most of the
PCs obtained from correlation coefficient and noise
distribution is generally less, unlike the unstandardized
components, facilitating more band combinations.
3. ALTERATION MAPPING
The Ariab Mininig District (AMD) of the central RSH is known
for gold mineralisation associated with silica-barite lenses and
dissiminated sulphides — (Wipfler, 1994). These
mineralisations are often surfacialy represented by
gossaniferous bodies of variable dimensions showing intense
oxidation effects. Magnetite is the most abundant oxidation
product observed in the massive sulphides (pyrite, sphalerite
and chalcopyrite). The country rocks display effects of
chloritization, sericitization, carbonisation, and silicification;
whereas high flourine content & the formation of hydroxylated
minerals added to the occurrance of alunite in the silica-barite
suggest involvement of hydrothermal solutions (Wipfler,
1994). Figure 4 shows a generalized geological map of the
Ariab area and the known gossaniferous bodies.
1 5
155 wr E
d
35 20 J:
[.] Quatemery cover RS nmin fli artzdtorite HI Granta granodiorhta
VA) eade volcanics WA Friciosticivoleanocinstic V2]. Awat Asoteriba Wl Gossan bodion aesociated
and Intrusives scie LS with massive aulphide
= Fault
Figure 4. Generalized gological map of the Ariab area showing the
Major gossaniferous bodies (modified after Wipfler, 1994).
3.1 Feature Oriented PCA
The application of PC transformation for mapping alteration
Zones has been introduced through a technique called
Feature Oriented PCA by Crosta & McM Moore (1989). In
this approach four input bands, out of which the desired
273
target has characterstic contrasting spectral features in two of
them, are selected and processed. The eigen matrices
contributed by these two bands are then examined for their
significance in last two higher-order components. The
desired target is often mapped in one of these PCs with
opposite signs. Details of the method has been well
elaborated in Loughlin (1991).
Two four-band sets: TM 1, 3, 4 & 7, and 1, 4, 5 & 7 covering
the Ariab sub-scene (area “A”, fig. 1) of the RSH (path/row
171/46) were selected to represent iron-oxide and hydroxyl-
rich minerals, respectively. The data were then geocoded,
atmosphericaly corrected and principal component
transformations performed. The 1st PCs mapped albedo
and topographic information whreas the 2nd PCs display the
VIS/IR versus SWIR bands in contrasting signs as often the
case is. The eigen vectors of PCs 3 & 4 were checked if
they contain significant loadings in opposite signs from input
bands of 1 & 3 and 5 & 7, as these band pairs are expected
to display contrasting response for iron and hydroxyl-rich
a)
TM1 TM3 TM4 TM7
PC1 0.38 0.60 0.53 0.46
PC2 -0.07 -0.24 -0.43 0.87
PC3 -0.84 -0.06 0.52 0.18
PC4 0.39 -0.76 0.51 0.08
b)
TM1 TM4 TMS TM7
PC1 0.33 0.46 0.71 0.42
PC2 -0.30 -0.69 0.22 0.62
PC3 0.84 -0.25 -0.39 0.27
PC4 -0.30 0.50 -0.55 0.60
Table 3. Eigen vector loadings of the Ariab sub-scene; a) for "iron-
oxide" mapping using bands 1, 3, 5 and 7, b)for "hydroxyl" mapping
using TM bands 1, 4, 5 and 7.
minerals, respectively (Loughlin, 1991). In both of the
considered cases PC4 appears to have mapped the iron-
oxide (-0.76) and hydroxyl-rich (-0.55) minerals in dark pixels
(table 3), and to obtain them as bright pixels these images
have been negated. This operation excludes the effect of
vegetation from both PCs by mapping them in dark pixels.
The next important step was to filter the images by 3X3 low-
pass matrix in order to supress the obviously incorporated
high noise. The effect of noise is apparently well pronounced
in the “hydroxyl” image possibly due to band-5 that also has
the largest variance and overall reflectance. A “mixture”
image was also produced by linear combination of the two
images. A composite image has been finally obtained by
using the "hydroxyl", "mixed" and "iron-oxide" imges in R, G
and B, respectively. Figure 5 shows the "iron-oxide" and
"hydroxyl" images before and after fileting.
Comparison of the the obtained composite with available
map (Wipfler, 1994) has shown that all the known
mineralisations have been mapped in a distinct yellowish
color against the bluish-red country rocks (Kenea, 1996). It
is also possible to suggest other potential sites which may
require field verificastions. Furthermore, it is to be interpreted
that the yellowish hue suggests the dominance of
hydroxylated-silicates in the mineralisation, and could be
followed also along most slopes close to the insitu altered
rocks.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
