International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
aerial photos and SI images, and some sparse trees. In the man-
made object area, the accuracy becomes worse. Here 2.4 % of all
points points have more than 4 meters difference. These points
are almost all positioned on the border of buildings and trees.
This may be caused by the fact that manually measured points
and automatically extracted points are referring to different
object points due to small errors in orientation procedure and 3D
modeling problems.
4.2 IKONOS Image Dataset, Thun, Switzerland
The test covers the area around the town of Thun, Switzerland. It
is about 17 x 20 km. The terrain elevations range from 600 m to
2200 m, with some very steep and high cliffs.
One IKONOS stereo pair was acquired on 2003-12-11 10:29
GMT over an area of approximately 11 x 20 km?. Another
IKONOS triplet was acquired on 2003-12-25, 10:39 GMT over
area of approximately 11 x 20 km”. The two areas have nearly
50 % overlap. All images are IKONOS Geo products. The
sensor and sun elevation angles (ca. 19 degrees) were less than
optimal. The low elevation angle of the sun causes strong
shadows, especially in the south part of the images and in
general low contrast images. A 2 m resolution reference DSM
generated from airborne LIDAR in the year 2000 was obtained
from the Swiss Federal Office of Topography, Bern. The
reference DSM only covers the south part of the whole area.
£e
fo ; = X vd T
Figure 7: Extracted DSM
Top: The whole area; Bottom: Town of Thun
Firstly, the original 12 bit images were preprocessed and resulted
in enhanced images for further image matching. Then the images
were orientated with the help of about 40 GCPs measured by
GPS. The orientation accuracy is about 0.41 m in planimetry and
0.68 m in height. Finally, a 5 m raster DSM was generated from
the matched mass points and the edges. Some areas like lakes
and rivers are defined as “dead areas” manually.
Figure 7 shows the 3D visualization of the generated DSM (the
whole test area and part of the town area). The results show that
even small geomorphological features are extracted and surface
discontinuities are well preserved.
132
Table 2: DSM accuracy numbers (west part of test area).
areas; C-City areas; V-Tree areas; A-Alpine areas.
RMSE | * >
(m)
Mean
(m)
Compared m 50m m
Points
Area
+
+
Table 3: DSM accuracy numbers (east part of test area).
O-C areas; C-City areas; V-Tree areas; A-Alpine areas.
»
m 50m m
Mean | RMSE
Compared
(m) (m)
Points
Table 2 and 3 give the DSM accuracy test results. We compute
the differences between the interpolated heights from our DSM
minus the heights of the reference DSM. The accuracy of the
generated DSM is at 1.3 — 4.8 pixel level. It depends on the
terrain type. Higher accuracy can be achieved in open areas. In
urban and tree areas the accuracy becomes worse. The analysis
shows that points with more than 6 m differences are almost all
distributed in the tree and urban areas. In open areas (with some
sparse trees and small cluster of houses), more than 70 percent of
the points have differences of less than 1 meter. The results show
significant biases. This is caused by different point definitions in
laserscan and photogrammetric surface points. Also, the different
acquisition times of the IKONOS images and the laserscans may
play a role . For more details see (Eisenbeiss, et al., 2004).
in meter
Table 4: Reference DEMs. Height accura
5x5 5km x 5km
5x5 5km x 5km
5x5 5km x 5km
5x5 5kmx
25 x 25 10km x 1.3km
25 x 25 1 x 7.7km
50km x 30km
z. x“
ed DSM (the whole test area)
Figure 8b: Reference DEM (left, 25 meter grid of dataset “DLR-DEM-
05-2”) and the generated DSM (right, 25 meter grid)
4.3 SPOT5 HRS Image Dataset, Bavaria, Germany ;
Here we report the work carried out within the ISPRS-CNES
Initiative on DEM generation from SPOTS-HRS stereo images.
For details see (Poli et al., 2004)
The test area (No. 9) covers a part of South Bavaria and a part of
Austria (approximately 120 x 60 km’). Table 4 gives
information on the reference DEMs. A stereo pair from SPOTS-
HRS was acquired on 1 October 2002 in the morning. The
International Arc.
INC idu e
images have the
across-track direc
The test area incl
inclined alpine ar
inclined areas). O
a large number o
The TIN based
points and the edg
Figure 8 shows tl
results show that t
the references, b
because of the 5m
Tables 5 and 6
orientation accura
height. We comp
reference DEM :
Table 6 shows th
tree areas manuall
From Tables 5 an
e The accuracy «
Ipixel level or es
about 2 pixels, b
trees.
o All datasets still
failed to detect.
e The results show
biases are about 1
significantly nega
‘are higher than tl
expected..
Tab
In this paper we
approaches for fi
array images witl
matching strategy
matching with nei
edge matching. TI
little or no texture
contribution of ob
to refine the ma
accuracy. The ge
specific sensor mc
the rigorous sens
RF (Rational Func
As evidenced by
reproduce even s
from the quantita
concept leads to e
buildings is taken
pixel or even bet
SPOT) as “best c
aerial images (foc
one pixel rule" c