the appli- 
lue of this 
'sents the 
was gen- 
dge about 
he recon- 
able con- 
enerating 
indprinzip 
tereowin- 
en Unter- 
spaketes 
lemodells 
uivalente 
viederum 
| definier- 
ufgaben- 
endungs- 
The stereo angle is an important parameter for the deriva- 
tion of digital terrain models (DTM) from remotely sensed 
image data. It characterizes the different viewing angles 
(Fig. 2). 
a - stereo angle 
  
Fig. 2 Stereo angle 
As reliable experimental investigations are either too 
expensive or simply impossible, the value for this parame- 
ter mainly is only estimated. Our investigations indicate 
that the use of a simulation technique may be the best way 
to get a satisfying result. 
Such a simulator has been developed in the Institute of 
Space Sensor Technology (Reulke, N., 1995; Reulke, R., 
Reulke, N., Jahn, H., 1994). It allows the generation of 
image data of a very complex scene, encompassing the 
radiating or reflecting surface of a planet, the propagation 
of the radiation through the atmosphere and through the 
optical components, the conversion of the optical into an 
electronic signal, and the signal processing. 
The simulator consists of three parts. The first one simu- 
lates the geometric properties (like orbit, planet, position of 
the sun), the second one includes the radiometric condi- 
tions, and the last module takes care of the analogous 
electronic part of a camera (Fig. 3). 
  
scene (planet, orbit, sun) 
ray tracer 
Geometry | : ; | 
intersection points with DTM 
  
  
  
  
atmosphere, reflections 
Radiometry radiance transport 
radiance at the sensor 
  
  
  
  
  
charges 
signal processing 
digital image 
Camera 
  
  
  
Fig. 3 Simulation tool 
Algorithms for the data processing can be added. The 
application of this simulator gives the opportunity to opti- 
mize camera parameters and to define optimal detecting 
conditions. 
The basic idea was that an optimal stereo angle strongly 
depends on the structure of the observed terrain. With the 
following procedure our simulation technique makes possi- 
ble such an optimization: 
With the help of the simulator image data have been ge- 
nerated, basing on a digital terrain model obtained from 
previous Mars missions. A DTM of the Mars was chosen, 
because WAOSS is a Mars-camera. Besides there should 
be no limitation regarding other landscapes, the DTM was 
changed by different exaggeration factors for our investi- 
27 
gations. 
These images were matched in order to find conjugated 
points. A match algorithm of the Technical University of 
Munich was used (Heipke, C., 1994), which evaluates the 
local image point correlation. The first step is choosing a 
number of start points. Outgoing from these start points 
the matcher tries to find conjugated points. The result of 
this procedure is a file containing the coordinates of the 
conjugated points in the image plane. 
The knowledge of the location of the conjugated points, of 
the simulated flight position and of the viewing direction of 
every detector element allows the generation of a new ter- 
rain model. This part of the processing cycle yields a more 
or less dense net of elevation samples. Elevation values 
between these samples are available after an interpolation 
procedure. 
By comparing both DTM’s a error criterion can be defined. 
The analysis of the spatial and temporal dependencies of 
this criterion on any parameter of the environment and/or 
the camera then enables the adaptation of a camera 
parameter to the relevant scientific task (Fig. 4). So an 
optimal parameter can be found. 
  
  
real DTM Simulator 
  
  
  
  
  
  
  
  
  
  
received Matcher & 
DTM 3D-Recon- 
  
struction 
  
  
  
  
  
  
Fig. 4 Scheme of processing 
2. OPTIMIZATION OF THE STEREO ANGLE 
Already from the beginning it was predictable that a few 
processes should influence the optimal stereo angle. The 
most important ones will be outlined. 
The relative shift of a certain point in two images is the 
decisive measure for the determination of its elevation. 
Since there is an absolute error due to the spatial discreti- 
zation by the optoelectronic system, the relative error 
decreases with an increasing stereo angle (Fig. 5). Accor- 
dingly, the whole DTM must become more accurate. 
o - stereo angle 
  
Fig. 5 Relative shift of conjugated points in dependence on 
different stereo angles 
The opposing effects are: The larger the stereo angle, the 
bigger the perspective distortions. It means that different 
CCD-lines get different impressions of one area (Fig. 6). 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996