speckle
ies like
1 SAR
errors,
tes like
around
ccuracy
20 data
rate a
(DEM).
for the
pective
'e was
relation
n data
' holes,
ork. An
tify and
ope of
nally, a
om the
h size.
te the
ure 3 in
graphic
ailable,
ps in a
| 4. For
dicated
t of the
to the
nted in
ndicate
] bright
ference
4. For
of 47.8
| been
Figure 3: Stereoscopically derived DEM.
Figure 4: Map derived reference DEM.
Figure 5: Height differences between map-derived and
stereo-derived DEM.
4.2. Data Geocoding
Based on the parametric mapping models and on the
map-derived as well as the stereo-derived DEM JERS-1
image data were geocoded. An intercomparison of
geocoded images produced with map-derived and stereo-
derived DEM, respectively, was made in order to further
conclude on the usefulness of the stereo-derived DEM.
The respective results are presented in Raggam et al.
675
(1995). In this context only the geocoded OPS-2 and the
geocoded SAR-2 image are comparatively presented in
Figure 6 in order to show the global information
characteristics of JERS-1 image data.
Beside the JERS-1 data, a geocoded multispectral SPOT
image as well as an ERS-1 SAR image are shown in
Figure 7 to demonstrate the radiometric but also the
geometric differences to the JERS-1 images. While the
optical data present themselves very similar, it can be
seen for the SAR data that the shapes of the layover
areas are completely different and much more extended
in the ERS-1 data. Consequently, for mountainous terrain
the content of useful information in ERS-1 images is
significantly reduced in comparison to JERS-1 SAR
images.
Moreover, for data from the steep looking ERS-1 sensor it
happens more frequently, for instance, that extended dark
patterns occur inside a geocoded layover area. Such
patters arise already from small error effects and in fact
this kind of geocoding errors is less severe than
anticipated from the visual impression.
5. CONCLUSION
For the high alpine testsite Ótztal, a multisensoral image
data set comprising optical images from JERS and SPOT
as well as SAR images from JERS and ERS-1 was used
for stereoscopic investigations and for the production of
geocoded images. Based on the achieved results the
following conclusions can be made:
e From a radiometric point of view optical JERS data
provide a good performance for stereo mapping, as a
stereoscopic image pair is acquired during one
overflight. In this concern JERS data are superior to
SPOT stereo pairs being collected in separate
overflights and with a certain temporal difference.
e From a geometric point of view the JERS stereo
disposition is limited to a stereo intersection angle of
some 15 degrees. The resulting base-to-height ratio of
about 0.3 is insufficient to obtain a high accuracy in
height, in particular in comparison to image pairs from
the SPOT sensor with respective values of even more
than 1.0.
e SAR stereo images of the JERS sensor in general
provide a worse potential for stereo mapping, as the
stereo intersection angles are very small. In
comparison to ERS-1, however, the JERS sensor
produces less layover. This is of high benefit for the
stereo mapping task, because the areas excluded from
successful image matching are significantly smaller.
e Sometimes, systematic errors may be resulting from
stereo mapping, as expressed by mean height errors
deviating significantly from 0. These effects have to be
further investigated. In this concern, other correlation
algorithms may be helpful.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
