From a study of formula (31) it becomes evident that the feasibility of 
the proposed solution depends on the possibility of inverting a more or less 
large matrix of normal equations. In special applications of photogrammetry, 
such as laboratory measurements, ballistic measurements, terrestrial appli- 
cations and, more recently, cadastral surveying, using airborne photography, 
the single model still plays & dominant role. The total number of unknowns 
in such cases may very well be within the limits which today can be reduced 
by inverting the corresponding normal equation matrix. However, depending on 
the number of relative control points considered essential in order to satisfy 
the requirement for redundant information and on the type of the electronic 
computer available, there will arise in many photogrammetric applications the 
problem of treating numbers of unknown parameters exceeding the computational 
capacity for reduction by direct inversion. For such cases, it is desirable, 
however, to maintain the advantages of the systematism and simplicity described 
earlier in forming the observational equations and the corresponding normal 
equations. Consequently, the solution to our problem must concern itself with 
methods for determining the roots of a normal equation system of the type shown 
in formula (31). 
B. A Solution by Partitioning 
In order to further reduce the number of unknowns in the normal equation. 
system obtained in formula (31), we split the B matrix and the A vector in 
  
Such & way that one group is associated with the model and the other group with 
the camera orientations. We denote the corresponding submatrices and subvectorg 
by Bx»Bo and Ay and Ay , respectively, With these notations we can 
present the system of normal equations (31) as follows: 
[B CaP""AT Y! Bax [B CAP" A Y' Bo]ao s BL CAP AY B G9) 
C A the a EE Rr Re ABERHAN SE NIRE AA AU EY ENUAR AR Ro A SR oru RReREER ROLE £D 
24 
ERE LE