More
ation
nodel
ically
ation,
ation
ation.
esent
eling
jects
class
these
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Digital Elevation Model
Gradient
Figure 2. DEM and Gradient image. The higher
elevation/gradient, the darker a pixel is. The elevation is in the
range (0, ) m.
2.0 Data
The A.S.T.E.R. image captured on 22/09/2001 with ID pg-
PR1A0000-2001092201 019 057 was used. It covers the area
geographic coordinates (38.3235, 20.5307), (38.2266, 21.2362),
(37.6733, 21.0737), (37.7694, 20.3734). The bands 01 (green),
02 (red), 03 (near infrared) of the V.N.I.R. sensor were used
with spatial resolution 15 m (Fujisada, 1998; Kahle et al.,
1991). The radiometric conditions (gains) for it's band during
the data acquisition was high gain for bands 01 and 02, normal
for band 03 (Abrams and Hook, 2002; Chavez, 1996)
R->03, 6->02, B->01
Figure 3. Color composite image
Additionally a digital elevation model (DEM) was used (figure
2). Contour lines were digitised from a topographic map
produced by the Hellenic Military Service. The map scale was
1:50000 and the contour interval was 20 m. A DEM with grid
size 30 m was derived by interpolating the contour lines. The
gradient is in the range (0, ) degrees (figure 2).
495
2.3 Pre-processing of A.S.T.E.R. Imagery
First radiometric correction was implemented. Each band was
distriped by using the histogram matching technique. Then
digital values were converted to radiance on the basis of the
corresponding band gain. Finally path radiance was removed on
the basis of linear regression of green and red band to the near
infrared band (Eastman, 1999).
Then non parametric geometric correction was implemented by
the use of a second degree polynomial and ground control
points (gcps) derived by field survey with a hand-held Garmin
G.P.S. The root mean square error of the transformation was
16.47 m for 14 gcps. The images were resampled to 30 m. A
color composite image of the corrected bands is given in figure
Figure 4. Cluter Map
2.4 Cluter Map
Training areas were selected for the major thematic classes
occurring in the study area (Table 1).
ID | Thematic Class Occurrence
(number of pixels)
1 Fir Forest 9,660
2 Barren 24,023
3 Cultivated 1 22,757
4 Cultivated 2 21,714
5 Cultivated 3 20,609
6 Cultivated 4 3,307
7 Lake 329
8 Mixed Forest 86,242
Table 1. The thematic classes and their occurrence
in the cluter map.
Maximum likelihood classification (Mather, 1987) defined the
landcover classes of the study area. The cluter map derived is
given in figure 4 while the occurrence of it's class is given in
table 1 and figure 5.
2.5 Geomorphometric description of cluter map
Each thematic class was described on the basis of mean
maximum and standard deviation of both elevation and gradient
(table 2 and table 3) (Evans, 1980; Florinsky, 1998; Mark,
1975).