od of the united model construction can be explained in the same manner 
as that in Arrangement B (See Figs.-5a and 5b). In the modified technique 
the first, second, and third pictures have the potential to provide five 
interior orientation unknowns, and the potential of the second, third, and 
fourth photographs determines seven interior ones. Thus, the first united 
model is similar to the object, while the second united model is not sim- 
ilar to the object. Then, the geometrical characteristics of the self cal- 
ibration problem may also be shown as in Figs.-5c and 5d. 
Arrangement D: 
Suppose that two metric and two non-metric cameras are configurated in such 
a way that the first and fourth photographs are taken with the metric cam- 
eras and the second and third pictures by means of the non-metric cameras 
(See Fig.-6a). The self calibration with these four overlapped photographs 
can easily be performed if the same part of the object has been imaged in 
common on the first and fourth pictures (See Fig.-4b). ‘However, when it is 
not the case, the modified method for the united model construction may be 
employed. The first, second, and third photographs have the potential to 
provide seven interior orientation unknowns. Thus, the relationship be- 
tween the first united model and the object can be determined by 10 orien- 
tation parameters (seven exterior and three interior). The same discussion 
is valid for the process of the second united model construction with the 
second, third, and fourth photographs. Connecting these two united models 
  
  
Fig.-6: The self calibration problem for the case of Arrangement D of two metric and two non-metric pictures. 
- 568 - 
A 
era 
LA ^ P £^ bert