the areas that previously had few or no measurements. On the
other hand, the choice of points with not very good texture led to
more blunders. Our blunder detection method rejected 4% of the
points, and another 1.8% of the points were rejected by a second
method for detection of DTM spikes based on robust statistics.
The remaining points showed slightly more remaining errors
than the first DTM.
S. ORTHOIMAGE GENERATION
Using the first DTM and the ground to image PMFs an
orthoimage was derived from the fore channel. Its planimetric
accuracy was checked as explained in the previous section by the
use of the GCPs and the parallaxes between the orthoimages of
the fore and aft channels. Another possibility, that was not used
here, is to derive new GCPs by finding corresponding points in
the two orthoimages, projecting back in the original images and
finding ground coordinates through a forward intersection. This
procedure works even if the DTM used for the orthoimage
generation is erroneous (see Baltsavias, 1996).
The time required for the orthoimage generation (2.4 Mbytes) on
a Sun Sparcstation 20 was 39 sec. For an orthoimage of the whole
fore or aft channel ca. 3.5 min are required. An overlay of the
orthoimage on the DTM is shown in Figure 1.
6. CONCLUSIONS
The data set used for this test is limited. In addition, problems
with the GCP identification and the high image noise make this
test a bit more difficult than what should be expected with normal
quality imagery. However, the great precautions that we took in
the measurement of the image coordinates of the GCPs and the
image preprocessing lead to the conclusion that the accuracy can
not be much better with “normal” imagery. Even under these
conditions, we proved that by using the fore and aft channels with
a simple, fast but strict sensor model needing only ca. 10 GCPs,
an accuracy of 6 - 7 m in planimetry and height can be achieved.
When using the nadir and one of the fore or aft channels, the
planimetric accuracy should be higher, but the height accuracy
should stay, due to the worse B/H ratio, at more or less the same
level.
Automatic DTM generation with a novel matching algorithm that
makes use of geometric constraints and has no problem in
matching of images with any scale or rotational differences was
performed. Due to lack of extensive reference values no
definitive conclusions on the DTM accuracy can be drawn. Based
on the available qualitative and quantitative measures the RMS
error of the DTM raw data is 0.5 - 1 pixel, with maximum error
close to 30 m. These values will of course vary depending on the
form and coverage of the terrain. A dense regular DTM grid will
however be less accurate in areas with few or no measurements
due to poor texture. To fill-in these gaps and to correct for the
few, relatively smal,l remaining errors a postediting is required.
Orthoimage generation poses no problem and can be fast. The
only important requirement is a good quality DTM. The
planimetric accuracy that was achieved was in the order of half a
pixel.
116
Future investigations will make use of the nadir channel for
evaluation of the point positioning accuracy, and DTM and
orthoimage generation. In cooperation with the University of
Melbourne we will use the roving GPS data for DTM evaluation.
Furthermore, new tests using the planned MOMS-Priroda images
over areas with good reference DTMs will be performed.
Acknowledgements
The authors would like to acknowledge the assistance and
cooperation of Prof. C. Fraser and his colleagues, University of
Melbourne, who provided data and information on the Australian
testfield, Dr. W. Kornus, DLR, who provided the images,
approximate image coordinates of the GCPs, and information on
the MOMS-02/D2 system parameters, and Dr. V. Kratky who
made modifications to his computer program to make data
processing easier and more accurate.
References
Ackermann, F., Bodechtel, J., Lanzl, F., Meissner, D., Seige, P,
Winkenbach, H., 1990. MOMS-02- A Multispectral Stereo Im-
ager for the Second German Spacelab Mission D2. In: Interna-
tional Archives of Photogrammetry and Remote Sensing, Vol.
28, Part 1, pp. 110 - 116.
Baltsavias, E.P., 1991. Multiphoto Geometrically Constrained
Matching. Ph. D. Dissertation, Institute of Geodesy and Photo-
grammetry, ETH Zurich, Mitteilungen No. 49, 221 p.
Baltsavias, E.P., 1996. Digital Ortho-Images - A Powerful Tool
for the Extraction of Spatial- and Geo-Information. ISPRS Jour-
nal of Photogrammetry and Remote Sensing, (in press).
Baltsavias, E.P., Stallmann, D., 1992. Metric Information Ex-
traction from SPOT Images and the Role of Polynomial Map-
ping Functions. In: International Archives of Photogrammetry
and Remote Sensing, Washington D.C., USA, Vol. 29, Part B4,
pp. 358 - 364.
Baltsavias, E.P., Stallmann, D., 1993. SPOT Stereo Matching for
DTM Generation. Proc. of SPIE, Orlando, USA, Vol. 1944, pp.
152-163;
Ebner, H., Kornus, W., Ohlhof, T., 1992. A Simulation Study on
Point Determination for the MOMS-02/D2 Space Project Using
an Extended Functional Model. In: International Archives of
Photogrammetry and Remote Sensing, Washington D.C., USA,
Vol. 29, Part B4, pp. 458 - 464.
Fraser, C.S., Shao, J., 1996. Exterior Orientation Determination
of MOMS-02 Satellite Imagery. Geomatics Research Australa-
sia, No. 64.
Fraser, C.S., Fritsch, D., Shao, J., Collier, P.A., 1996. Ground
Point Determination Using MOMS-02 Earth Observation Im-
agery. Presented Paper, 37th Australian Surveyors Conference,
Perth, April 15 - 19.
Kratky, V., 1989. Rigorous photogrammetric Processing of
SPOT Images at CCM Canada. ISPRS Journal of Photogramme-
try and Remote Sensing, Vol. 44, pp. 53 - 71.
Seige, P., 1993. Status of the MOMS-02 Experiment on the Spa-
celab Mission D2. Proc. of ISPRS Workshop "International
Mapping from Space", Hannover University, pp. 39 - 50.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
KE
Reg
app
cad
by :
resi
pro
Die
Pro
mit
elat
ein
Reg
sick
Af
sat
ten
ma
mo
to
Lar
acc
ven
pro
tal
thi
thi
Wis:
