»ments only. 
be refined, giving 
ore important we 
ood measures to 
strategy in its last 
ometric stochastic 
nizing roads and 
In: Automatic 
from Aerial and 
1, O. Kuebler, P. 
d extraction by 
nal Archives of 
g (IAPRS). Vol. 
Computer Vision. 
1986. Extracting 
tions on Pattern 
Vol. PAMI-8 N.4, 
ation and Visual 
ire extraction and 
ind Image 
| Organization in 
oposal for a 
'ansaction on 
3. N.2, pp.382- 
Mayer, B. Radig, 
from images. In: 
de Objects from 
by A. Gruen, O. 
rlag, Basel. 
tracing by profile 
. In Automatic 
from Aerial and 
1, O. Kuebler, P. 
)96 
EXTERIOR ORIENTATION DETERMINATION OF MOMS-02 THREE-LINE IMAGERY: 
EXPERIENCES WITH THE AUSTRALIAN TESTFIELD DATA 
Clive Fraser and Juliang Shao 
Department of Geomatics 
The University of Melbourne 
Parkville, Victoria 3052, Australia 
Email: Clive Fraser8mac.unimelb.edu.au 
jsha@sunrise.sli.unimelb.edu.au 
Commission III, Working Group 1 
KEY WORDS: Three-Line Satellite Imagery, Triangulation, Accuracy 
ABSTRACT 
The three-line imagery recorded by MOMS-02 on its inaugural mission in 1993 has 
provided an opportunity to evaluate the photogrammetric potential of such satellite 
imaging systems. Among the areas covered by MOMS-02/D2 three-fold stereo imagery was a 
110 km x 40 km swath in Central Australia which has been selected as an accuracy 
testbed for ground point determination and DTM extraction. This paper first discusses 
the establishment of the Australian Testfield, both in respect to photogrammetric 
requirements and the collection of ground truth data by GPS survey. The triangulation 
accuracy of MOMS-02/D2 imagery is then considered. A review of the mathematical model 
is provided and the results of a series of bundle adjustments employing different 
control point configurations, numbers 
of orientation images and orders of 
interpolation functions are analysed. The findings of the accuracy evaluation are 
discussed and compared to theoretical expectations. 
1. INTRODUCTION 
with “the ‘launch in April 1993 of the 
Modular Opto-electronic Multispectral 
Stereo Scanner, MOMS-02, as part of the 
German Spacelab Mission D2 on board the 
Space Shuttle, the photogrammetric 
community was presented with a first 
opportunity to evaluate the metric 
potential of a space-borne high- 
resolution  three-line imaging system. 
MOMS-02 was specifically designed to 
generate digital topographic mapping 
data, "primarily! through automated  DTM 
extraction to accuracies of about 5m, and 
through the generation of digital 
orthoimagery with adequate resolution and 
geometric accuracy to meet map 
specifications at 1:50,000 scale and 
larger (e.g. «Ackermann et al, 1990; 
Seige, 1993; Fritsch, 1994). 
Simulation studies (Ebner et al, 1992) 
had indicated that object point 
triangulation accuracies to better than 
5m in planimetry and up to 5m in height 
should be possible from the along-track, 
three-fold coverage of MOMS-02. Imagery 
obtained from the D2 mission would 
provide verification or otherwise of 
predictions. o£. this metric performance 
level. 
One of the Mode 1 panchromatic imaging 
sequences recorded during the 10-day D2 
mission, namely scene 17 of Orbit 75b, 
covered a swath of 110km x 40km in the 
north eastern region of Central 
Australia. Through a collaboratıve 
arrangement between the Department of 
Internati 
onal Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
Geomatics at The University of Melbourne 
and MOMS -02 photogrammetric research 
teams in Germany, this area was chosen as 
an ‘accuracy testbed’ for exterior 
orientation determination and ground 
point triangulation. The main requirement 
for the ‘Australian Testfield’ was the 
provision of a well distributed array of 
image-identifiable ground control points 
which would facilitate a comprehensive 
metric evaluation of MOMS-02/D2. three- 
fold stereo imagery. 
The purpose of this paper is to report on 
two facets of the overall investigation 
into the cartographic potential of MOMS- 
02, namely the establishment of the GPS- 
surveyed accuracy testbed and the 
investigation into ground point 
triangulation accuracy attainable with 
this high-resolution (HR) imaging system. 
2. THE AUSTRALIAN TESTFIELD 
The testfield covered mostly flat and 
featureless terrain which displayed an 
elevation range of only 70m. One. of the 
primary requirements for the  MOMS-02 
image testbed was the availability of 
image-identifiable ground control points 
which could be accurately surveyed by 
‘GPS. In this regard the testfield left a 
little to be desired. Image identifiable 
points were effectively restricted to the 
dams and track/fence intersections of the 
Lake Nash cattle station. To achieve 
triangulation accuracies at the 5-10m 
level, sub-pixel image mensuration 
precision is required. This in turn is 
207