13
This apparent conflict can be easily clarified by the fact that
the system of equations 4.08 can actually be formed stepwise in a form
of summation. By virtue of equation 3.0 3 the term (C^WOj^ for the
object point, Oj, appearing in m photographs can be evaluated from
(assuming that is a diagonal matrix):
(C* W OT 1
m
l
i = l
.... 4.10
which takes the maximum size of 3 x 3. Once 4.10 is computed, the matrix
U for Oj is readily evaluated from 4.09 as:
Uj = [W - (W C) (C 1 W C)“ 1 (C* W)]j
• • • «
4.11
Finally, with reference to equation 3.04, the entire system of normal
equations of 4.08 may be obtained from:
l (B* U B) . D = - l CB t U F). .... 4.12
j=l 3 j=l 3
It is now evident that the scheme presented above is well suited
for electronic computers and that the maximum size of the matrix to be
inverted is 47 x 47. Such an inverse usually constitutes the major opera
tion on which depends the size of computer to be used.
The only instrumental operation in analytical photogrammetry is
the measurement of plate coordinates on a comparator. The major problem in
this operation is the difficulty in identifying the object points to be
used in the solution. This difficulty often leads to inconsistent plate
coordinates which, when used as input for sub-block relative orientation,
affects the internal accuracy of the model. One practice to overcome this
difficulty is to examine the residuals at each of the used points and those
displaying large values are eliminated and the solution repeated. To avoid
the necessity of performing the entire operation, including the main matrix