channels are not arranged perfectly parallel on the focal
plane plane (Fig. 15). Hence, matching of images obtained
by different CCD lines (oriented in an oblique direction with
respect to each other on the fictitious focal plane) should
produce disparity data with a similarly oblique trend of the
matched pixels relative to the original referenc
pixels.
In Figure 16, the results for automatic matching of parts of
the stereo 1 and 2 channels to the nadir are shown for a
selected image line. The nadir channel is the reference
channel and is not plotted in the diagram (it can be thought
as a horizontal line at a difference of 0). It is evident from
the plot that the pixel positions on the stereo 1 and stereo 2
CCD lines have distinct trends towards increasing and
decreasing offsets to the pixel positions in the nadir
channe, respectively, as is expected from the orientations of
CCD lines in the focal plane (Fig. 15). Hence, these data
provide a crude qualitative check on the geometric
calibration.
The test also shows that the matcher achieves ist design
goal to obtain disparity data with subpixel accuracy.
o -
£e
— |
21.0 pe ÀÓ« —
3 i
o
B s $o: 5
3 $ © -
= ut ee |
“0.5 ; -
o : x
= d
3 4 A iss
20.0 CHEN ai
E P. Ir x "
3 F : X x =
gp : Gig
20.5 — x
o - x -
5 o
a ER: x stereo 2 : : -
o NITE NIN ew NO al a Te 00
200 300, 400 3500 . 600. 700 4800
pixel (sample) number in reference image
Figure 16: Pixel number of reference image (nadir
channel) vs. difference between line number in the
matched images (stereo channels 1 and 2, respectively)
and the reference image. The scattering of the plotted
values around the observed trend is attributed to
atmospheric turbulences caused by large temperature
gradients over the surface of Lake Constance (see
Discussion).
5. SUMMARY AND DISCUSSION
The HRSC camera which is going to be launched to Mars in
November 1996 was operated successfully in the first
outdoor test. The results demonstrate that HRSC, for the test
cases that were studied, fully meets the design goals.
The instrument operates properly (pixel binning, change of
scan rate) according to camera commands. Image data in the
panchromatic channels are highly reproduceble. The
expected effects of the camera rotation on the brightness
levels are being confirmed by the test data. The flat field
files determined during radiometric calibration appear to
correctly describe the different responsitivity of sensor
elements. The correction significantly improves the image
quality. Compression/decompression artifacts are visible,
but negligible. They do not appear to affect the further
354
processing, e.g. the flat-field correction and image
matching. We also carried out a successful rough
verification of image scale and resolution and verified the
general properties of camera metric properties and the
geometric calibration data.
Finally, ET3 also provided an important verification of the
flight software. We verified the correct operation of parts of
the preprocessing software, the decompression code, the
flat-field correction and the software that computes time tags
for each image line. All software modules were found to
fulfill the requirements. The digital image matcher, an
important cornerstone program of the photogrammetric
processing, works according to the design goals. Large
contiguous image areas can be "mapped" and pixel
coordinates of conjugate points can be determined at sub-
pixel level. It is noted that scattering of the disparity data
are seen in the images. However, we carried out independent
tests using the Gotcha matcher (Day et al., 1992) which
gave similar results, indicating that the scattering does not
represent artifacts generated by the matching software.
Instead, we think the scattering at sub-pixel level to be due
to atmospheric turbulences caused by large temperature
gradients over the surface of Lake Constance.
More thorough analyses of the test data are currently under
way, as at the current state of this study and within the
limited time available, no comprehensive, but only sporadic
checks could be done on the huge amount of image data that
was collected. However, for complete analysis of the
performance of HRSC and the ground data processing
system, more sophisticated tests must be conducted. For
this purpose, an airborne experiment using the HRSC flight
spare model will be carried out near Mount Etna in Sicily,
Italy, later this year. The goal of the experiment is to
acquire a data set which can be subjected to a full
photogrammetric analysis, resulting in large-scale digital
terrain models (DTMs) and color ortho image mosaics.
Acknowledgements: We wish to thank our co-workers T.
Roatsch, G. Schwarz, C. Reck, and the Dornier Test Team
who provided much help in the acquisition and processing of
the ET3 data. We also wish to thank our collegues at the
Technical University of Berlin for important software
contributions for this study.
6. REFERENCES
Day, T. et al., 1992. Automated Digital Topographic
Mapping Techniques for Mars. In L.W. Fritz and J.R. Lucas
(Ed.), International Archives of Photogrammetry and
Remote Sensing, 29(B4), pp 801-808, Washington D.C:
American Society of Photogrammetry and Remote Sensing.
Neukum, G. et al., 1995. The Multiple Line Scanner Camera
Experiment for the Russian Mars 96 Mission: Status Report
and Prospects for the Future. Photogrammetric Week ‘95,
pp. 45-61, Heidelberg: Wichmann, 1995.
Oberst, J. et al., 1994. Mars “94/96 Pushbroom Cameras:
Plans for Ground Data Processing and Analysis. Proceedings
of the Symposium “Mapping and Geographic Information
Systems”, ISPRS, Vol. 30(4), pp. 533-540.
Oberst, J. et al., 1996, Photogrammetric Analysis of
Clementine multi-look-angle images obtained near Mare
Orientale, Planet. Space Science, in press.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
pro
cla:
pre
sen
cou
