In: Wagner W., Szflcely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
558 
different input datasets (IARR, PCA and MNF) in identifying 
the environmental impact of the salt plugs. Bands 1-9 of 
ASTER in combination with Principal Component Analysis 
(PCA), Minimum Noise Fraction (MNF) transformation and 
Multi-Layer Perceptron (MLP) were used in this study. 
2. STUDY AREA 
The study area (28° 31' - 28° 53' N ; 52° 16' - 52° 33' E) is 
situated in the Zagors fold-and-thrust belt, western the Iranian 
province of Fars, southeastern Shiraz, and about 25Km south 
west of Firouzabad (Fig. 1). The Zagros mountain range is 
divided into three tectonic zones from the NE to the SW: the 
High Zagros, the Zagros Simply Folded Belt, and the Zagros 
Foredeep Zone (Stocklin 1968; Falcon, 1974). The study area is 
located in the Simply Folded Belt (SFB) which has particularly 
been studied owing to the salt plugs and its structure. The 
geology consists of Inffacambrian diapirs (salt plugs) 
surrounded by the Cretaceous to recent formations. 3 
Figure 1. Geological map of the study area, Southern 
Firouzabad, SE Shiraz, Iran 
3. METHODS 
VNIR+SWIR dataset of ASTER were used to detect and map 
salt plugs-affected areas by the MLP neural network. ASTER 
instrument measures reflected radiation in three bands between 
0.52 and 0.86 pm (VNIR) and six bands from 1.6 to 2.43 pm 
(SWIR), with 15- and 30-m resolution, respectively (Fujisada, 
1995). The ASTER Level IB data used in this study were 
acquired on March 24, 2001. The following steps constitute the 
data processing and analysis of the ASTER bands: (1) spatial 
registration of the 30-m SWIR data to the 15-m VNIR data; (2) 
the data were geometrically corrected using 1:25000 
topographic maps; (3) Internal Average Relative Reflectance 
(IARR) calibration was then carried out on the data to 
normalizing images to a scene average spectrum. This method 
is particularly effective in areas where no ground measurements 
exist and little is known about the scene (Kruse, 1988); (4) A 
spectral reduction and data compression was performed using 
the principal components analysis; (5) To train and validate the 
use of MLP networks, training areas of each lithological unit 
were selected using knowledge of the PCA and the geological 
map. To do this, several ROIs were measured and extracted 
from ASTER image. 
3.1 Principal Components Analysis (PCA) 
Principal components analysis (Richards, 1984; Eklundh and 
Singh, 1993) has become a standard statistical approach in 
image processing for two main reasons: (1) to reduce the 
number of correlated image bands to form a small number of 
independent principal components to represent most of the 
variability carried by the multiple image bands, and (2) to 
increase the interpretability of the components as combinations 
of multiple bands (Jing and Panahi, 2006). PCA output results 
were used to create RGB color composite images to 
discriminating various lithological units and reducing the 
information included in the raw data into two or three bands 
without losing significant information (Monger, 2002). 
Component 
Eigenvalue 
Variance (%) 
Total (%) 
PCI 
0.2735 
87.989 
87.99 
PC2 
0.0238 
7.661 
95.65 
PC3 
0.0094 
3.044 
98.70 
PC4 
0.0016 
0.520 
99.22 
PC5 
0.0012 
0.388 
99.60 
PC6 
0.0005 
0.173 
99.78 
PC7 
0.0003 
0.108 
99.89 
PC8 
0.0001 
0.060 
99.95 
PC9 
0.0001 
0.052 
100.00 
Table 1. PCA statistics of VNIR-SWIR ASTER bands on 
study area 
PCA statistics were accounted to selecting components with the 
highest information to be used in selecting training areas. 
Table 1 shows the eigenvalues, variances and total cumulative 
variances for the nine PC image of ASTER data. The PCI 
image shows 87.99 percent of variances. The PC2 and PC3 
images show 7.66 and 3.04 percent of variance respectively. 
Therefore the first three components represent 98.7% variances 
of the image data. On the other hand components 4-9 only 
contain 1.3% of the information. 
In order to mapping lithology and environmental impact 
assessment of the salt plugs components 1-3 were used to 
generate colour composite image and to select training areas. 
3.2 Minimum Noise Fraction (MNF) 
The MNF transformation is a linear transformation related to 
principal components that orders the data according to signal- 
to-noise-ratio (Green et al., 1988). It can be used to determine 
the inherent dimensionality of the data, to segregate noise in the 
data, and to reduce the computational requirements for 
subsequent processing (Green et al., 1988; Boardman and 
Kruse, 1994). The MNF was applied to the ASTER to enhance 
lithological units and salt plugs-affected areas. 
3.3 Multi-layer perceptron (MLP) 
The multilayer perceptron (Rumelhart, and MacClelland, 1986) 
is by far the most well known and most popular neural network 
among all the existing neural network paradigms. (Hu and Neng 
Hwang, 2002; Carvalho, 2001). It is a mathematical approach 
(Hu and Weng, 2009), with some advantages and disadvantages 
as compared with other existing neural networks. For example, 
nonparametric statistical methods may be more useful for