‘about 4 um or % 
). 
ts, reveals that the 
about +35 um (2 
in object space. 
e control points, it 
nation remained in 
yrmation is caused 
n the left and the 
e curves available 
lower and upper 
ows). 
ind the right side 
the control points 
“0.8 mm in object 
rtainty of definition 
f configuration a) 
  
ve | r.m.s. error 
S in image 
) 3.8 um 
nfiguration a) 
  
  
  
  
of configuration b) 
   
  
  
  
  
  
No. tie No. nodes No. curve r.m.s. error 
curves points in image 
40 224 6554 4.3 um 
  
  
  
Table 4: Adjustment results of configuration b) 
As mentioned above, the accuracies achieved with 
configurations a) and b) are quite similar; however, 
configuration a) has some advantages regarding the 
computational effort. Thus, in practice, it is advisable to 
use rather low bent tie curves. 
Having a configuration of low bent tie curves one might 
be attempted to represent the curves by parabolas 
instead of joined cubic polynomials. As mentioned 
above, the tie curves of configuration a) initially have 
  
  
been described by parabolas. Adjusting this initial 
configuration resulted in r.m.s. errors at the control 
points in the order of 44 um or 3 pixels. As described, 
this result has been improved decisively (see table 3) 
by inserting additional nodes in order to refine the tie 
curve's shape. So, even low bent lines of the car are 
better represented by free-form curves than by 
parabolas. 
After having finished image orientation a model of the 
car can be reconstructed. The object curves are 
obtained by spatial intersection of free-form curves. 
Image orientation is regarded to be constant during this 
process, so that long and flexible object curves can be 
modelled (see figure 11). 
c S s ud 7 A 
B - x 
  
Figure 11: Object curves reconstructed from the finally oriented images 
5 CONCLUSION 
Applying photogrammetric object reconstruction 
techniques we can build models from spatial objects 
using images as source of information exclusively. By 
means of the method presented in this paper, real 
world objects having no or hardly any point information 
but enough line features can be dealt with. This could 
be advantageous, for instance, if targeting is not 
possible. Additionally, complex free-form object curves 
extracted from digital images might be a valuable 
support for automated detection of homologous 
features, thus increasing the level of automation in 
digital photogrammetry. 
References 
Forkert,.....G.,. 1993. Photogrammetric Object 
Reconstruction using Free-Formed Spatial Curves. In: 
Gruen, A. and Kahmen, H. (editors.): Optical 3-D 
Measurement Techniques Il. Herbert Wichmann 
Verlag, Karlsruhe, 1993, pp. 221-228. 
  
International Archives of Photo 
grammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
Forkert, G., Dorffner, L., Klimpfinger, H., Rusch, W., 
1994: The Application of Visualization Techniques for 
Diagnosis and Quality Control. ISPRS Comission Ill 
Symposium, Munich 1994, pp. 194-200. 
Forkert, G., Kerschner, M., Prinz, R., Rottensteiner, F., 
1995: Reconstruction of Free-Formed Spatial Curves 
from Digital Images. In: IAPRS, Vol. 30, Part 5Wl, 
ISPRS Intercomission Workshop “From Pixels to 
Sequences“, Zurich 1995, pp. 163-168. 
Kager, H., 1980: Das interaktive Programmsystem 
ORIENT im Einsatz. International Archives of 
Photogrammetry and Remote Sensing XXIII/B5, 
Hamburg 1980, pp. 390-401.. 
Kager, H., 1989: ORIENT: A Universal 
Photogrammetric Adjustment System. In: Gruen, A. 
and Kahmen, H. (editors.): Optical 3-D Measurement 
Techniques. Herbert Wichmann Verlag, Karlsruhe, 
1989, pp. 447-455.