pe EE d 
DETERMINATION OF A DIGITAL TERRAIN MODEL FROM SPACELAB METRIC CAMERA 
PICTURES BY MEANS OF DIGITAL IMAGE CORRELATION 
Frank Boochs 
Rudolf Decker 
Institute for Photogrammetry 
University of Bonn, West Germany 
ABSTRACT 
The presented article shows an example for the determination of a digi- 
tal terrain model (DTM) from metric camera pictures using the correla- 
tion program ARCOS. Two images (N° 885, 887) were digitized by a high 
precision microdensitometer. For the control of the correlated heights a 
digital terrain model was available. The comparision of the correlated 
with the reference heights was accomplished in three separate point 
raster with about 2000 points in total. The attained accuracy was exami- 
nated in three categories to differentiate between characteristics which 
might have influence onto the result (i.e. contrast, topography, object 
type). Finally the achieved point quality was checked with respect to 
mapping purposes. 
1. INTRODUCTION 
In the end of 1983 the first metric camera experiment was started. On 
board of spacelab a modified Zeiss RMK 30/23 rotated in an orbit of 
about 250 km height above the earth and took the first series of metric 
photographies from space. The campaign was designated to provide pictu- 
ral data for mapping or interpretational purposes. The two main charac- 
teristics like high resolving power and geometrical accuracy of the 
images made the data interesting for the applications in question. As 
first investigations have shown /Guyenne, Hunt 1985/, the data is of 
high value for photogrammetric processes, although the image quality 
didn't reach the expected level. With further improvements (forward 
motion compensation, for example) this type of spaceborne date should 
become an important tool for the future. 
Due to the promising first results it is to expect, that this type of 
data might be extensively used for mapping of great parts of the earth. 
Then it would be interesting to have simple and automated means for the 
height determination process. One possible aid might be the image corre- 
lation technique, for which an example is given in this article. 
The applied correlation procedure is based on the program package ARCOS, 
which produced encouraging results for medium scale photos /Boochs, 
1984a/. As the surface topography and the image structure for spacelab 
images are quite different from standard medium scale photographies, it 
is interesting to evaluate how the correlation procedure reacts and 
which accuracy will be attained. /Decker, 1986/ has done very intensive 
investigations onto these questions and some results will be presented 
here. 
The paper is divided into five chapters. The first one describes the 
data charecteristics, the second one shows the preparations for the 
- 79 -