Grün: 
  
Förstner: 
Grün: 
Ehlers (USA): 
  
Ackermann: 
  
  
transfer test, is probably due to the fact that 
the operator selected well suited image areas with 
good texture, at points which would normally have 
been selected for point transfer. This may not be 
an average case, representing overall image resol- 
ution. 
It looks like Dr. Fórstner has a direct response to 
this problem. 
We have a small test field with targeted points 
which we photographed with the same camera as we 
use in a Planicomp but without using the film and 
without using the measurement equipment of the 
Planicomp. The direct pixel coordinates of the 
targets were put into a bundle block adjustment 
and with high redundacy we got sigma nought of 1.3 
micron with a pixel size of 27 which shows that 
the pixel size is really not limiting the precision 
to a - let's say - to a tenth of a pixel. 
The next, and I think the last question comes from 
Dr. Ehlers. 
My question goes into the same direction. I have 
no doubt that digital correlation is powerful for 
several kinds of applications, and in particular 
for aerial triangulation. My question is - do you 
have any estimation about the influence of your 
hardware in particular the A/D conversion error on 
your precision? 
We have not done very thorough investigations. The 
internal precision of image correlation was 1.1 and 
1.2 microns. After block adjustment, however, we 
came to equivalent values of about 3 microns. Most 
likely that difference is mainly due to hardware. 
The Planicomp itself may have one and a half micron 
mechanical and other errors. The second contribution 
is certainly the calibration of our CCD cameras and 
of their interfacing with the Planicomp system. 
We have used both systems alternately. I think the 
questions of calibration and of hardware are more 
Significant, at the moment, than the question of 
pixel size. 
Thank you very much. Unfortunately I have to cut 
off this discussion right now. I guess, we have 
Some more discussion on similar points coming up 
in the next Sessions. Just one remark. I feel 
the need of processing on-line, point positioning 
on-line in the object and in the image space, is 
the stronger the more we switch from images digi- 
tized on analytical plotters to fully digital 
systems. Anyway, it was an interesting discussion. 
I would like to thank you all. 
56