66
From the transformations on the fiducial marks we have already derived the
influence on the accuracy of a) the lack of flatness, b) the irregular film shrin
kage, c) the operator and d) the transformation itself. This part of the standard
error of unit weight was 15.55 jum as an average.
When we now compute the discrepancies between the transformed image co
ordinates and the corresponding values from the test object, we expect a larger
standard error of unit weight after adjustment, because it will include the uncer
tainties of the co-ordinates of the test object. The average image scale for one
plane is approx 37: 22 and for the other 37 : 32. Assuming the standard error
of the test object to be o g , the resulting standard error in the image scale will be
1 /(37 / 22) 2 + (37 / 32) 2
n„ / = 1.44 o g •
The root mean square of the standard errors of unit weight of the actual
adjustments is 30.8 ¡urn. We can now estimate the accuracy of the test object
from the following relation
30.8 2 = 15.55 2 + 1.44VV
og = 22.1 ¡urn
This value is highly significantly different from the estimated standard de
viation of the final co-ordinates of the test object, which was 4.0 ¡urn. See chap
ter 8.3.2.
If we assume the different determinations of the interior orientation to be
uncorrelated (which may not be true) and perform an analysis of variance, we
cannot detect any irregularities of the interior orientation among the different
determinations. A dependency between the different determinations of an ele
ment may be introduced by the possible constant errors of the given co-ordina
tes for the points in the test object.
The final values of the elements of interior orientation of the apparatus are
given on the last lines of Table 13. They can be regarded as constant and then
the apparatus is well suited to photogrammetric measurements.
8.4. AERIAL CAMERAS
The method for calibrating cameras, that is described in this thesis, is pri
marily intended for close-up cameras. However, it would be very interesting
to check the interior orientation of aerial cameras under operational conditions
using three-dimensional test fields to get information not only on radial distor
tion and accuracy but also on principal point and principal distance. However,