ng 
nd 
wn 
ile 
all 
ini 
es, 
'ed 
ent 
tic 
are 
ral 
are 
  
  
for 
ied 
the 
on. 
to 
to 
ace 
n a 
leir 
the 
ue, 
ue. 
ms 
are 
> is 
mean brightness values (8-bit scale) 
Band 1 Band 2 
1 
sos 
MALALLLALLLLLULLLLVALLLABLRALALLR LAUT 
  
FUMATATA TITAN SAE 
aA 
"n 
ALAR 
^ni 
Band 3 Band 4 
1 
ne 
| 
un 
FLL 
FIATALLALLLILLL UL LL LA TASALARAAA LAA SLLALARA LR 
  
BALELELELLAUN 
CRAAMRAAAXAARARA 
MAUR LAURE 
m 
Band 5 Band 6 
  
aL 
QRANAXARASARAANSRARARARARASARAXRNA TAN 
  
  
MALLILLLBLELKAKLELLLLLTALT cuves 
nae i 
n 
Figure 2. Structure of Brightness Values ( T (v) 
3.2 Suitability of Bands 
Brightness values of several points corresponding to the 
bands of both TM and JERS were read and their mean 
values plotted to find the spectral response 
characteristics. These are presented in Figure 3 (TM 
bands only). The response figures vary considerably from 
band to band which is encouraging. There are 6 bands of 
JERS and 7 bands of TM giving us a reasonable choice 
of bands. 
mà Lai Ce A 
«= o c os x: oe o 
o o o o o c a o 
sv 
c 
  
e 
  
e 
NA 
4 6 8 10 12 
wavelength 
Figure 3. Mean Reflectance Against Corresponding 
Wave Length (bands of TM) - Expansive Soil 
(Sample 1) 
3.3 Laboratory Analysis of Specimen 
The minerals responsible for giving expansive charac- 
teristics to the expansive soils are generally found buried 
beneath the surface. In rare cases however they are 
found to be exposed to the surface. Some of them have 
been collected by the second author and four specimens 
were subjected to laboratory analysis for knowing the 
spectral characteristics. One of the response curves 
appear in Figure 4. The laboratory result could not be 
incorporated in the signature generation operation. 
33 
  
130% 
120% 
110% 
  
  
  
405 492 67 71 871 972 1073 
REFL. WAVELENGTH (na) 
Figure 4. Spectral Characteristics of specimens 
4. TRAINING SAMPLE 
Extensive site investigations have been carried out in 
Muscat for construction purposes and for locating places 
suitable for development. Four such areas were selected 
as training samples. The sample patches vary in size; 
one of them, rectangular in shape, measure about one 
sq. km. Boundaries of these patches were surveyed 
using a hand held Global Positioning System (GPS). 
They were identified on the satellite images with the 
help of the GPS coordinates. For this the images had to 
be rectified beforehand using Ground Control Points 
(GCP) located and read on the available topographical 
map of the area. The mean values of the residuals 
indicate reliability of the rectification achieved : 
OE =18.3 m ON =24.8 m 
The GPS was used in single observation mode which is 
less accurate than the double observation mode. So this, 
together with the rectification error, contributed to the 
location uncertainty. But considering the fact that with 
limited number of drilling there is always an uncertainty 
of exact location of the boundary and further noting the 
fact that expansive soil is not a sharply defined land 
cover it can be said that sufficient accuracy has been 
achieved. 
S. CLASSIFICATION 
Classification accuracy depends upon the data quality, 
data structure (distribution type), number of features, 
noise in the data, the number of bands employed, 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996