bul 2004
s in the
ught to
nt eg,
-grazing
| various
nges in
ng and
data in
sses and
e area,
llow the
ynmental
egies of
=
^
satellite
and its
' remote
images,
it 2) and
sat 4).
Facility
2/35 and
led from
bove).
path 172,
Acquired
image,
9. 4788 ,
sat Ltd.
1967 and
38-2 and
»resented
satellites.
one, the
onboard
| (MSS),
Mapper
rs which
'nces are
esolution
h spatial
>]. There
e end of
e size i$
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
considerably less than with Landsat. Also the radiometric
resolution is higher than before (DigitalGlobe, 2004).
The CORONA satellites, used 1960 — 1972, were the first
generation of US reconnaissance satellites and they took stereo
photos using two oblique viewing panoramic cameras. One
photo is a panchromatic film strip of 7 x 90 cm. The best spatial
resolution is 1.83 m for the KH-4B mission at a flight height of
150 km. The SYGIS project has utilized photographs from the
KH-4A mission, the size of the photos being 2.25 x 29.8 inch
and the resolution 2.7 m (www.earthexplorer.usgs.gov).
Geometric correction of the CORONA images is very difficult
due to the lack of information about the principal point and
fiducial marks, and panoramic distortions (Altmaier - Kany,
2002).
3.2 Digital elevation model
A digital elevation model (DEM) covered part of the study area
and its data sources were SRTM-DEM and ASTER-DEM. The
DEM was used to remove shadow areas from satellite image
interpretation results and to make 3-dimensional visualizations
(still images and flight movies) of the ground.
3.2.1 SRTM-DEM: The Shuttle Radar Topography Mission
(SRTM) provides a DEM at resolution levels of 30 and 90 m
covering the earth between latitudes 60N and 57S measured in
February 2000. The DEM is constructed using synthetic
aperture radar (SAR) interferometry, meaning that two radar
images have been taken from slightly different positions and the
surface height is determined using phase differences between
images. The DEM was measured using two frequencies (C- and
X-bands). X-band data was used in this study. The drawback is
that the swath width is 45 km, meaning that there are gaps
between neighboring orbits. WGS84 is used as horizontal and
vertical datum. This means that ellipsoidal heights are provided.
The vertical accuracy should be better than 16 m absolute and 6
m relative, and <the horizontal accuracy better than 20 m
(Rabus et al., 2003). The SRTM-DEM was provided by DLR
(The German Aero-Space Center) to SYGIS (the Finnish project
in question) as a project member of the DLR project.
3.2.2 ASTER-DEM: ASTER (Advanced Spaceborne Thermal
Emission and Reflection radiometer) is a multispectral imager
which was launched onboard Terra satellite in 1999. ASTER
has 14 spectral bands from visible to thermal infrared region
and their spatial resolution varies from 15 m to 90 m. There is
also one telescope looking backward in the near infrared region
(channel 3B, 0.78 — 0.86 um, spatial resolution 15m) to give
stereoscopic capability in the along-track direction. Digital
elevation model generation algorithm uses only instrument and
satellite ephemeris data. The geodetic map projection is WGS84
and pixel spacing 30m. The horizontal and vertical accuracy
should be better than 50m and 15m, respectively (ASTER,
2002). Two ASTER-DEMSs were acquired through NASA Earth
Observing System Data Gateway
(http://edcimswww.cr.usgs.gov/pub/imswelcome/).
3.2.3 Matching of the SRTM- and ASTER-DEMs: SRTM-
DEM covered only part of the study area so it was decided to
augment the DEM by ASTER-DEM. The coverage of DEMs
can be seen in Figure 2., red represents the coverage of SRTM-
DEM and green and blue that of ASTER-DEMs. Because the
heights of SRTM-DEM are from the surface of ellipsod WGS84
(Rabus et al., 2003) and the heights of ASTER-DEM were
relative, there was a need to study the heights of different DEMs
more carefully.
899
Compared to STRM-DEM, the heights of ASTER-DEM 1 (the
light green small triangle visible on the top in Figure 2) were
46.9 m lower and root-mean-square-error (RMSE) between
DEMs was 48.5 m. The heights of ASTER-DEM 2 (the darker
blue small triangle visible below in Figure 2) were 22.8 m lower
than SRTM-DEM and RMSE was 27.0 m.
Figure 2. The study area and the coverage of DEMs, the largest
red area represents the coverage of SRTM-DEM and the green
above and the blue beneath ASTER-DEMs. Data copyright
DLR and Eurimage.
It was decided to adjust the ASTER-DEMs to STRM-DEM.
After removing the average difference between DEMs, it was
noticed that the slope of the linear regression equation between
SRTM and ASTER-DEMs were not equal to |, meaning that
there was scaling factor in the ASTER-DEMs which needed to
be removed. Therefore the adjustment of ASTER-DEMs were
based on the regression equations between SRTM- and ASTER-
DEMs. The correlation between dems were high, the correlation
coefficients were 0.9855 for ASTER-DEM 1 and 0.9952
ASTER-DEM 2. After the adjustment, the RMSE between
SRTM-DEM and ASTER-DEM 1 was 7.8 m and ASTER-DEM
2 10.6 m.
3.3 Georeferencing
The georeferencing of satellite images and digital elevation
models was performed by choosing the panchromatic channel of
Landsat ETM image as a master image and georeferencing
other images and DEMs to that master image using ground
control points. The master image was georeferenced using
ground control points measured from UK military aviation
maps (TPC G-4C, TPC G-4D, Scale: 1: 500 000, UK 1998).
Other images were georeferenced using an ETM-image. 20 - 30
ground control points were determined from images and first
degree polynomial transformation and nearest neighbor
interpolation was used in each case. The root mean square
errors were less than one pixel. The CORONA images were not
georefenced due to the previously mentioned difficult imaging
geometry.
The effect of georeferencing was studied by comparing the
georefenced images using orbital information and the
georeferenced images using ground control points. The corners
of the images were measured and errors computed between the
mean coordinates. Errors in east and north directions were
defined as the ETM-coordinate minus the corresponding
coordinate from other image and planimetric error is length of
error vector. Table 2 represents the results. The results indicate
