gets is between 
ie light for larger 
0—a) 
  
2 
popu 
get center 
ST) 
ptical axis. 
lane 
arget plane 
nage 
mikrons 
  
  
  
pb p HEE 
alpha (gon) 
iene 92m / 5mm 
——cb5m/5mm 
tM 5mj1qmm 
———— Sm / mm 
— 5m '/ 20mm 
—— 2m 7 OMIM 
— 2i / 6mm 
2m / 20mm 
  
  
  
Figure 6: Estimation of offset for a 25 mm lens and a maximum image radius of 18mm. 
5 Influence on photogrammetric bundle adjustment 
With the help of a simulation it will be shown which 
parameters of the bundle adjustment are influenced if the target 
is oversized. 
The simulated network consist out of 20 images from 5 
locations (Fig. 7). The object was designed with 51 spatially 
distributed object points. Targets are located in three different 
distances ( 25 points app. 4.6m, 25 points 2.6m, 1 point 1.6m). 
For the simulation parameters of a large format film camera 
/Dold and Suilmann 1991/ were used. This camera has an 
image format of 230mm x 230mm, a focal distance of 165mm 
and a typical image measurement accuracy of about 1 micron. 
For the simulation the target diameter was chosen in this way, 
that the offset in the convergent images from location 1 to 4 
were up to 5 microns (Fig. 8a). As explained above, no offset 
influences the images from location 5 because the image plane 
is parallel to the plane of targets (Fig 8b). With these offsets 
the true image coordinates were changed before they were used 
as an input for the bundle triangulation. Two different versions 
of bundle triangulation were calculated. Version 1 uses a free 
network adjustment without simultaneous camera calibration 
and version 2 uses a free network adjustment with 
simultaneous camera calibration. 
The result of both bundle triangulations show that the 
simulated offset will only slightly influence the residuals of 
image coordinates. The residuals of image coordinates are 10 
times smaller (less than 0.5 microns) than the simulated offset 
(Fig 9a, 10a). 
In cases with no simultaneous camera calibration the offset 
influences not only the parameters of exterior orientation 
(location and orientation of camera) but also the object 
coordinates (app. 20 microns deviation to nominal values Fig. 
9c and 10c ). In cases with simultaneous camera calibration the 
offset influences also the camera parameters. Thus the 
influence on the object coordinates becomes smaller (app. 10 
121 
microns). Remarkable is, that the influence of the offset is 
mainly compensated by the exterior orientation parameters 
(app. 8096) (Fig. 9b, 10b). Figure 9b and 10b show the 
deviation between the true target centers and those target 
centers which are the result of the projection of the target 
centers in object space into the image by using the changed 
exterior orientation parameters which were calculated by the 
bundle triangulation of version 1 (Fig. 9b) respectivly version 
2 (Fig. 10b). 
It is dangerous is that the systematic offset is not clearly visible 
from the residuals of image coordinates but all other 
parameters of the bundle triangulation change. The influence 
on the exterior orientation is not critical, but the influence on 
the camera parameters is especially dangerous in those cases in 
which the camera parameters were determined in a calibration 
network and will be than fixed for using in a different network. 
location 2 
location 1 i TT 
ETT location 5 | a-30 
a=30 = : 
| 
- : location 3 
location 4 i i 
| 
| 
  
Figure 7: Photogrammetric network. On each location four 
images were taken (rolled by 0°, 90°, 180° and 270° 
around the optical axis) 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996