- 4 -
taken with the camera. This can be done by taking a series of
pictures and by determining the interior orientation of each of them.
Using the formerly described calibration method we obtain the
standard errors of the elements of interior orientation for each of
the pictures. These standard errors describe the statistical
variation within the pictures, but from the values of the elements of
interior orientation for different pictures, we can compute an ex
pression for the variation between the pictures. If these two variat
ions are expressed as variances, it is possible to apply Fisher's
F-test. *) In this way we can test the hypothesis of constant interior
orientation. The accuracy of the interior orientation can be expressed
as standard errors by taking the square root of the variance within or
between the pictures. The variance within is used when the variance
between is not significantly larger than the variance within, otherwise
the variance between is used to estimate the accuracy.
A recording cinecamera with fiducial marks, used in a study of air
craft in flight, has been calibrated using the described steel cage. It
is a General Scient Corp Multidata IV No. 1045, with Miltar lens No.
G 1379. The focal length is 75 mm and the image size 24 x 24 mm,
i.e. a narrow opening angle. Four film sequences were taken with
apertures f: 16, f: 11, f:8 andf:5.6. From every sequence three
negatives were chosen, one at the beginning, one in the middle and
one at the end, 12 in all. The mathematical model to define the
interior orientation of the camera had five parameters:
Two coordinates for the principal point,
the camera constant, c
two radial distortion parameters, a^, a^, which express the
radial distortion
3 3
dr = a_r + a_r .
3 5
The six ordinary parameters for the exterior orientation were also
introduced into the adjustment of the observations. The computations
are programmed for a CD 3600 computer. The result contains among
other things coordinates of fiducial marks and principal point,
camera constant, radial distortion, standard errors of these values
and the standard error of unit weight. Some numerical results can be
seen in Fig. 1.
Fiducial Marks
The variation of the coordinates of the fiducial marks between
pictures within the same aperture has a variance of the mean co-
2 2
ordinate of s = 5. 16 mm . The variance of the means between
. 2 2
apertures is s = 23. 08 mm . The numbers of degrees of freedom