s x =s z = 1,22x2,5=3 micron
(30)
Series of tests of the standard error of unit weight within photographs have
indicated that this standard error increases with the radial distance from
the principal point, see Hal/ert 1961 a and Hallert 1961 b. This is to be ex
pected because of the inevitable discrepancies between the central projection
of the imaging procedure and the orthogonal measurements. These discrep
ancies are primarily caused by lack of flatness of the image surface. This
surface can never be made exactly flat even if glass plates are used. The emul
sion will always have a certain variation in thickness, and the positions of the
details within the emulsion vary to a certain extent with the intensity of the
illumination. Moreover, flatness can never be measured free from error, and
the methods and procedures for measurement of flatness are in many cases
of little value as are also the expressions for flatness.
There are several determinations of the standard error of unit weight of
image coordinates available, see Schmid 1961, Rosenfield 1961, Brown 1959,
and Hallert 1960 b. In most cases no attention has been paid to possible varia
tion with change of radius. Therefore the available figures must be interpreted
as a certain mean value of the standard errors of unit weight. In the existing
publications and unpublished investigations by the author, it has been deter
mined that for glass plate cameras, c=150mm, image format 18x18 cm,
and measurement by skilled operators in the best available comparators (stand
ard errors of about 1 micron in each coordinate direction), realistic values
of the standard error of unit weight of image coordinates are of the order of
magnitude 2—3 microns in each coordinate direction. A certain variation
from the image center outwards must always be assumed. From experiment
it seems probable for the camera type mentioned that the range 1,5—3,5
microns can be expected.
Much more research should be devoted to the determination of this basic
standard error of unit weight and to the reasons for its variation. These in
vestigations must be made under actual operational conditions, and the
method of least squares must be used for the computations. It is most suitable
to combine calibrations of instruments and cameras with determinations of
standard errors of unit weight to distinguish between regular and irregular
errors.
Assuming a mean value of the standard error of unit weight of image
coordinates of about 2,5 microns for the camera with c=150 mm and an inter
polation within four stars as presented here, the mean value of the standard
error of the corrected image coordinates of an arbitrary object (well defined
in the image) is, according to expression (29),