SEMI-AUTOMATIC DERIVATION OF DIGITAL. ELEVATION MODELS FROM 
STEREOSCOPIC 3-LINE SCANNER DATA 
M. Lehner (+), 
R.S. Gill (+ +) 
( +) Deutsche Forschungsanstalt für Lu. und Raumfahrt (DLR), Oberpfaffenhofen 
(+ +) Computer Anwendung für Management GmbH (CAM), München 
ABSTRACT 
DLR is engaged in several stereo scanner projects 
(MEOSS on Indian Remote Sensing mission IRS-1E, 
MOMS-02 on German shuttle mission D2, HRSC and 
WAOSS cameras on Russian Mars mission). Besides 
fabricating the flight hardware for MEOSS camera DLR 
developed a software package for the evaluation of 
stereoscopic scanner data consisting of the following 
main subsystems: 
Image matching software for automatically locating 
large numbers of conjugate points in 3-line stereo 
scanner imagery. It is based on areal matching in 
image pyramids and subsequent local least squares 
matching for subpixel positioning. 
Photogrammetric adjustment software is used to derive 
digital elevation models and exterior orientation 
parameters. This combined point determination soft- 
ware is mainly based on collinearity equations for con- 
jugate and ground control points. Further it allows for 
the constraining of tne exterior orientation of the cam- 
era by given flight path and attitude data and by GauB- 
Markov statistical processes. 
The paper reports on the successful application of the 
software to airborne data taken with the satellite model 
of the MEOSS camera, including a comparison to an 
already existing DEM. Empirical and theoretical errors 
are given for various numbers of ground control points. 
Keywords: 3-line stereo scanners, image matching, 
photogrammetry, digital elevation models 
1. Introduction 
DLR is engaged in stereo line scanner cameras since 
1981, when the DLR Institute of Optoelectronics pro- 
posed to participate with a 3-line stereo camera 
(MEOSS: Monocular Electro-Optical Stereo Scanner) in 
an Indian satellite mission «Ref. 17. Parallel to the 
manufacturing of the hardware, work on data analysis 
started with simulation of triple stereoscopic imagery, 
establishing image matching software and testing the 
photogrammetric adjustment process via simulation 
« Ref. 2-47 . Since 1986 an airborne model of the cam- 
era was available. This was the first time thai a digital 
elevation model could be derived from such imagery 
« Ref. 5,7 7». MEOSS is now scheduled for a flight on 
Indian satellite mission IRS-1E end of 1992. Airborne 
test data of the satellite flight model have been evalu- 
ated for this paper. 
68 
In contrast to MEOSS which is a single optics system, 
the stereo scanner MOMS-02 (Modular Optoelectronic 
Multispectral Scanner) being built by the German com- 
pany Messerschmitt-Bólkow-Blohm (MBB) uses various 
optics of different focal lengths to take multispectral 
and stereoscopic images of the earth. This offers the 
possibility to derive digital elevation models and multi- 
spectral orthophotos of fine resolution for GIS-like 
applications. A description of the MOMS-02 scanner 
system can be found in <Ref. 6 >. This scanner wili be 
part of the payload of the German mission D2 now 
scheduled for a spaceshuttle flight in January 1993. 
One aim of the MOMS project is the development of a 
stereoscopic digital photogrammetric workstation in 
cooperation with several universities. 
Other stereo scanner projects at DLR are the Mars- 
94/96 WAOSS and HRSC cameras <Ref. 11,12>. 
WAOSS (Wide-Angle Optoelectronic Stereo Scanner) is 
meant for global monitoring of Mars (pixel size on 
ground 150 m near periapsis) and for the derivation of 
global topography of Mars whereas HRSC (High Resol- 
ution Stereo Camera, 10-20 m pixel size) will allow the 
derivation of fine-grid digital elevation models together 
with orthophotos in several spectral channels. Both 
cameras are - like MEOSS - monocular instruments. 
The derivation of digital elevation models from 3-line 
stereo scanner imagery consists of two main tasks: 
9 extraction of a large number of conjugate points 
from the stereoscopic imagery (image matching) 
and 
. reconstruction of exterior orientation and calcu- 
lation of the ground coordinates of the conjugate 
points via photogrammetric adjustment (combined 
point determination). 
The location of conjugate points in triple stereoscopic 
line scanner imagery can be divided into the following 
steps: 
* selection of patterns suitable for digital image cor- 
relation in the sense of an interest operator (the 
well-known Fórstner operator is used) 
* selection of search windows corresponding to 
these patterns in the images of the other looking 
directions 
e. digital image correlation (area based matching) to 
locate the pattern within the search area: image 
matching on pixel level 
. subpixel refinement of the correlation result (done 
here by local least squares matching technique). 
The: 
imag 
and 
Itere 
gran 
eque 
Som 
knov 
for 1 
bias 
mate 
In hi 
mod 
large 
wate 
corre 
mati 
auto! 
digit. 
desit 
oper. 
The 
wind 
appri 
and 
least 
and 
point 
the s 
thres 
of m 
surro 
We fc 
lated 
dows 
Furth 
dows 
ness 
varia 
interf 
old fc 
mate 
the r 
menti 
thres 
reaso 
intere 
Norm 
by the 
multir 
only. 
this [ 
repla: 
other 
promi 
given