e 
d 
ty 
le 
je 
n 
n, 
n 
% 
of 
je 
tilled cameras 
with adapter 
rings and 
teleconverter 
  
  
Fig. 2: camera configuration 
The camera system very often is larger than the 
object itself. In addition the combination of the 
different optical components is not very stable. 
So it was more convenient to rotate the object 
infront of the camera than to rotate the camera. 
2.3 Video cameras 
Solid state cameras exceeding the standard 
tv-size of 768 * 512 pixels are very expensive. 
For economic reasons the standard CCD-cameras 
Phillips LDH 0460/21 with the Valvo chip NXA 
1011 have been used. The pixel size of 1Oum x 
8um leads to 6.1mm x 4.1mm chip size. 
The same configuration as shown in fig. 2 had to 
be used for getting a sufficiant base to distance 
relation. The necessary distance between the 
object and the lenses required a focal length in 
the range of 40mm to 120mm. 
  
camera | size accuracy relative jresolution 
[mm?]| in image | accuracy [tp] 
  
Rollei 
6006 |60x60| s2um |#0.3x10° 4200 
Rollei 
  
4 
  
  
  
  
  
  
3003 |36x24 +2um +0.6x10° 2100 
Phillips 
CCD 6.1x4.1| +1um |#1.7x10"" 380 
  
Table 1: comparison of the used cameras 
The cameras operating with film do have a much 
better resolution like the CCD-cameras, in addition 
also the accuracy in relation to the film format or 
sensor size is better up to 5 times (see table 1). 
But the CCD-cameras can be used on-line which 
is necessary for example in applications for 
industrial manufacturing. If the cameras are used 
for quality control, the result has to be available 
within a very short time. That means, the whole 
process has to be done automatical with image 
correlation and/or automatic object identification. 
The listed accuracy (table 1) may not be reached 
in any case, especially in macro applications 
there are large problems with the image quality 
caused by the very small depth of focus. 
3. Calibration 
The photogrammetric solution is based on the 
reconstruction of the bundle of rays by means of 
the image coordinates and the inner orientation 
including radial symmetric distortion and 
systematic image errors. Calibration certificates 
have not been available for the used systems. For 
the stereo microscope the basic relations of the 
geometry have not been known in advance. A 
calibration and examination of the geometric 
stability was necessary. 
A complete calibration is only possible with a 
threedimensional testfield. The accuracy of such 
a testfield has to be at least in the range of the 
aspired object point accuracy. After several 
tests copies of reseaus mounted on u-shaped 
metalic plates have been used. The reseau itself 
is defining points in one plane, the lower part of 
the bevel of the metalic plate has been used 
instead of height control points (see fig. 2). 
  
Fig. 3: threedimensional testfield 
The thickness of the metalic plate is depending 
upon the depth of focus. Plates with a thickness 
between 1 and 3mm have been used. The depth of 
focus together with the tilted view was the limiting 
factor. Unsharp points cannot be used, at first 
the pointing accuracy is limited and outside of 
the depth of focus the optical relations are 
different. 
Caused by the very small view angles the unknowns 
of the inner orientation are strongly correlated 
and the accuracy of the required elements are 
limited. The calibration and also the computation 
of the photo orientation is only possible by bundle 
adjustment. The Hannover program BUNOR, 
belonging to the program system BLUH has been 
used. Caused by the correlations it was not 
possibletodeterminetheinnerorientationtogether 
with systematic image errors. But this was not a 
problem because of the large area of tolerance of 
the inner orientation parameters. The systematic 
image errors determined in a second step are 
related to the fixed inner orientation of the 
same images.