of the transformations*we get % 2 = 1.062 which is an incredibly good value
compared to % 2 (95%) = 1.145 and % 2 (5%) = 11.070 (The homogenity is too
good, because of the errors of the so called given values). Also the scale
factors indicate the same errors, since the variation between the transformations*
is highly significantly lower then the corresponding value within. (F = 0.116).
Fig. 10.
A target of the basement
’’cage”. The sizes of the
targets vary in such a way
that they become equally
large in a picture photo
graphed from a point 5 m
in front of the wall plane.
8.1.3. Interior Orientation.
In this case the interior orientation is defined by five elements: two co-ordi
nates for the principal point, the camera constant, and two parameters for the
radial distortion. A determination of the radial distortion and the accuracy by
means of the grid method has shown that all observations in the adjustment can
be given equal weights, i. e. unity. See [30]. The results are shown in Tables 4
and 5. According to Bartlett’s test the variation of the standard error of unit
weight among the pictures is not significant.
The correlation coefficients a posteriori for the principal point and the ca
mera constant are given in Table 6.
We assume the elements to be uncorrelated and we study the variation
of the interior orientation for the elements independently. The camera constant
varies highly significantly when the camera is focussed at 10 m. The reason for
this can be that the setting of the index on the lens introduces an error. If this
error is included, the camera constant has a standard error of 0.074 mm. If the
camera is focussed on infinity by screwing the lens until it stops, the camera
constant does not vary significantly. The co-ordinates of the principal point do
^ (see formulas 8.3.1 and 8.3.2)
43