'anbul 2004
and well
of highly
N. Adam.:
velopments
the ISPRS
, Hanover,
OM
1, “SPOTS
ise, 21-25.
ote Sensing
jaillarin, C.
ge ground
2003, IEEE
osium
aluation of
g chain and
te Sensing,
| of digital
ner data”,
5, 1992
rlag Berlin,
of Digital
ings of the
rom Space
r 2001
rroeder, V.
ta to Image
nformation,
Schroeder,
. HRS two-
'0 Data in
" Congress,
irst Results
rived from
ngs of the
om Space,
D. Scales,
n Models
mission ll
AUXISPRS,
te Sensing,
ok", S-NT-
CER. Stereo
pp. 10-13,
DEM GENERATION BY SPOT HRS
K. Jacobsen
Institute of Photogrammetry and Geoinformation,
University of Hannover, Germany
jacobsen@ipi.uni-hannover.de
TS HRS DEM Generation from SPOT-5 HRS Data
KEY WORDS: satellite, orientation, matching, DEM/DTM, analysis
ABSTRACT:
The HRS (high resolution stereo) is an autonomous imaging system on SPOT 5. It has been designed especially for the generation of
digital elevation models (DEMs). The combination of a forward and a backward view do allow a stereoscopic imaging with a height
to base relation of 1.2 and just 90 seconds time interval. Some images, which are usually not accessible and used only by SPOT
Image for the generation of DEMs, have been made available for the SPOT HRS study team. The test area Bavaria has been
analysed.
Just based on the known view direction, the orientation of a HRS stereo pair has been adjusted with the Hannover program
BLASPO. The mean square Z-discrepancies at the control points of 3.9m corresponds to an x-parallax of 0.6 pixels, a realistic value
for points located in areas with sufficient contrast. Depending upon the area, between 80% and 90% of the points could be matched
successful by a least squares method. The determined ground points do describe a digital surface model (DSM) and not the requested
DEM. Points located not on the solid ground have to be removed. For the 17 million points of one HRS stereo pair this can be made
only by an automatic process. The Hannover program RASCOR removed approximately 3596 of the DSM points. Nevertheless in
dense forest areas several points still may be located not on the ground. Corresponding to this, in the forest areas a height shift
against the reference data could be seen. But also the relative accuracy in the forest is not so good like in the open areas caused by
limited contrast and the not visible ground. In the open areas a vertical accuracy in the range of 5m and a little better has been
reached with a slight dependence upon the tangent of the slope.
1. INTRODUCTION based on sketches the location usually close to road crossings
can be estimated; so not the optimal accuracy of positioning can
The high resolution stereo sensors (HRS) of SPOT 5 includes be reached.
two optics viewing forward and after in the orbit direction with
incidence angles of 23°, generating a stereo model with a height
to base relation of 1.2. With 12 000 pixels a swath width of
120km is available. In the orbit direction, the pixel size on the
ground is just Sm, so a standard scene with 12 000 x 12 000
pixels is covering an area of 120km x 60km. The smaller pixel
size in orbit direction has advantaged for the vertical accuracy.
Figure 1.
SPOT HRS camera system
Figure 2: colour coded DEM generated with the HSR data and
location of the evaluation areas
Two slightly overlapping HRS-models, located in the south-east The investigated area has a height ranging from 270m up to
of Germany, in the federal state of Bavaria, covering also parts 1850m above sea level. Approximately 20% is covered by a
of Austria, have been evaluated. As control and check points for mixture of smaller and larger forests. Also some lakes are
the scene orientation trigonometric points have been used. included. The main part is flat up to rolling, only small parts of
These points cannot be identified directly in the scenes, only the Alps are included (see figure 2).For the analysis of the
