GEOMETRIC POTENTIAL OF MOMS-02/D2 DATA FOR POINT POSITIONING, DTM AND ORTHOIMAGE 
GENERATION 
Emmanuel P. Baltsavias 
Institute of Geodesy and Photogrammetry 
Swiss Federal Institute of Technology (ETH) 
ETH-Hoenggerberg, CH-8093 Zurich, Switzerland 
Dirk Stallmann 
Institute for Photogrammetry, Stuttgart University 
Keplerstr. 11 
D-70147 Stuttgart, Germany 
Commission IV, Working Group 2 
KEY WORDS: Satellite, Three-line Sensor Modeling, Geometric Accuracy Analysis, Geometric Information Extraction, Matching, 
DEM/DTM, Orthoimage. 
ABSTRACT 
The paper presents investigations on the geometric potential of MOMS-02/D2 data using three images over a testfield in Australia. The 
sensor model employed is presented and the point positioning accuracy for different control point, model parameters, and image 
measurement versions is analysed. The achieved accuracy for the fore/aft channel combination was 6 - 7 m in all three coordinates by 
using only 10 control points and quadratic attitude rates. Automatic DTM and orthoimage generation was performed again based on the 
fore and aft channels with 13.5 m pixel size. DTM generation is based on matching using geometric constraints and being able to match 
images with any scale or rotational differences. The evaluation of the DTM and orthoimage accuracy was based on limited reference 
values. The available qualitative and quantitative accuracy measures indicate a DTM accuracy (without any manual editing) of 0.5 - 1 
pixel RMS and a maximum error of ca. 30 m. The planimetric accuracy of the orthoimages is ca. 0.5 pixel. 
1. INTRODUCTION 
The great majority of past and current satellite-based optical 
sensors employ linear CCDs scanning in a pushbroom mode. For 
stereo acquisition either across-track (SPOT) or along-track (J- 
ERS 1) has been employed. The latter method offers the 
advantage of near-simultaneous image acquisition, thus making 
stereo interpretation and mensuration much easier. For this 
reason, most planned satellite-based optical sensors will employ 
this stereo mode, some in addition to across-track stereo for more 
frequent revisiting. The along-track stereo is implemented using 
two or three linear CCDs, usually each with a separate lens 
system. Another tendency is the increase of the geometric 
resolution. Available data from civilian electro-optical sensors 
have a resolution of up to 4.5 m, while several systems with 1 - 3 
m resolution are planned to be launched starting from 1996. 
These developments offer new possibilities for mapping - in 
continuously increasing scales - and derivation of important 
products like DTMs, orthoimages and orthoimage maps, 
classification maps etc. especially for many countries where this 
information does not exist, is outdated or not accurate enough. 
MOMS-02 is a high resolution, along-track stereo, three-line 
imaging system. Its sensor description and modelling, a system 
overview and other information are given in Ackermann et al., 
1990, Ebner et al., 1992, and Seige, 1993. Here, only a brief 
overview will be given. MOMS-02 employs seven channels, four 
multispectral and three panchromatic. For stereo imaging the 4.5 
m resolution nadir panchromatic channel and the 13.5 m 
resolution fore and aft panchromatic channels are used. The fore 
and aft channels have a look angle of 21.5? with respect to the 
vertical, thus leading to a B/H ratio of 0.8. MOMS-02 has 
different acquisition modes. The images used in this test were 
acquired with mode 1, and have a dimension of 2976 x 8121 
pixels (fore and aft), and 8304 x 8121 pixels (nadir). A nadir 
image that covers the same area as the fore or aft channel actually 
110 
consists of three subimages, that slightly overlap (120 pixels) in 
flight direction. Thus, an image covers an area of 40 x 110 km? 
and 37 x 109 km? for the fore/aft and nadir case respectively. 
The paper presents investigations on the sensor modelling and 
geometric point positioning accuracy of the MOMS-02 sensor, 
and methods, including accuracy evaluation, for generating 
DTMs and orthoimages. 
2. TEST DATA 
The three images used (fore/aft/nadir) were taken in April 1993 
during the D2 Space Shuttle Mission. They depict a deserted and 
flat region in the southeast Northern Territory, Australia with 
very little vegetation and almost no cultural features apart from a 
few dams, tracks and fences. Figure 1 shows a 3D view of an 
image part which is typical of the region. These images were 
selected because they were the only ones where high quality 
ground truth was available. 
The ground truth mainly consisted of about 80 points that were 
measured with GPS with an accuracy of ca. 10 cm during two 
campaigns in 1994 and 1995. A big problem was the 
identification of the points in the images. The definition of many 
points, due to lack of other more suitable cultural features, was 
poor (see Figure 4). The sensor resolution was too coarse for 
some points (e.g. fences), while the temporal difference between 
image acquisition and survey (e.g. different water levels) caused 
problems in identifying some good image points. Details on the 
testfield and the GPS control point survey are given in Fraser et 
al., 1996. Additional ground truth included a 17 km profile along 
a fence line with height data every 1 m and an accuracy of 10 - 20 
cm. The profile was acquired with a roving GPS antenna with 
aim the quality evaluation of derived DTMs. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
  
  
Fig 
lin 
gr 
rig 
p)