polarstereo, latitude longitude. When GCP exists, we can
make the absolute DEM.
By using quality flags, DEM is able to be interpolated from
neighbor data, converted into a fixed value.
DEM provided by observation of multiple times is possible
to be upgraded to high accuracy data. Also DEM mosicing
is available.
5. ESTIMATION OF ACCURACY
5.1 DEM accuracy using ASTER simulation
When there is no error at orientation elements and stereo
matching, the precision of DEM is decided by B/H and
ground resolution.
As ASTER data weren’t available yet, we made ASTER
simulation data made from JERS-1 OPS data.
ASTER data is stereoscopic on the along track direction
same as JERS1data.
We estimated the accuracy of ASTER DEM data through
ASTER simulation data were processed.
The condition of process is follows.
(1) Original DEMis used to elevation data to be generated
from 1:25000 topographic map by Geographical Survey
Institute of Japan.
(2)Image data is resampled from JERS1 data.
(3)Orientation image data is calculated from JERS1 data.
(4)ASTER backword image data is calculated from original
DEM and orientaion image data.
The result of estimation shows table 2. Also figure 3
shows generated ASTER simulation data and figure 4
shows original DEM and generated DEM.
Figure 5 shows the result of extraction of water area as
impossible matching area.
Table 2 Accuracy of Elevation
Positive | Negative | Average |R.E.M.
Maximum | Maximum | Error (m) | S(m)
Error (m) | Error (m)
175 -199 0.87 12.31
As this case has no error in orientation data, the cause of
error seems to be mismatching.
The orientation elements are decided by system
configuration. The position knowledge of EOS-AM1
spacecraft becomes high to utilize data of TONES
satellite compared with others spacecraft. Therefore
orientation error isn't mainly cause in elevation error.
And, It is easy to be matching on stereoscopic image in
along track direction, compared with one on cross track
direction. Therefore, the accuracy of DEM using real
ASTER data will be not change one of simulation data.
876
5.2 The caurse of Mismatching
The area where mismatching are occurred are estimated
that JERS-1 data were processed about 30 scenes
include desert zone, tropical rain forest zone, a grassy
plain, steep mountain zone. As a result, the area where
mismatching are occurred had the next characteristic.
(1) when the same intensity level
(2) when resemblance pattern continues
(3) when saturation data exist
(4) when area of the sea, cloud level exists
(5) shadow department of clouds
JERS-1 data has the capability that are B/H=0.3, ground
resolution=18m and 24m. In case of JERS1 data, the
accuracy of DEM is settled 60m-80m. It's average is
estimated with 30m-40m.
The results of experiment were able to get several
characters near by a theory value, showed in the table.
On the other hand, ASTER data has the capability that are
B/H=0.6, the ground resolution=15 m. In case of ASTER
data, the accuracy of DEM is settled 12.31m. It’s near a
theory value.
6. FUTURE
ASTER DEM product's accuracy will be 12.5m that has
the potentiality of drawing 1:100,000 scale topographic
map.
The Algorithm will be upgrading, especially searching
method of stereo tie point, extraction method of
mismatching area.
7. REFERNCE
(1)ERSDAC, 1995. Algorithm Theoretical Basis Document
for ASTER Level-1 Data Processing.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996