10
A number of plates have been measured and a brief description of the results is shown below. All measure
ments have been made by the microscope method which does not require the surface to be touched and
which gives high accuracy in the measurements. See reference 1.34:1. From comprehensive precision and
accuracy tests with different operators the following data have been obtained.
Precision. From replicated and repeated settings the standard deviation of one setting was found to be
s = 1.8 /¿m. Through replicated or repeated settings and determination of the average, the precision
can be increased as the standard deviation of the average decreases linearly with the square root of the
number of replications or repetitions.
Accuracy. From test measurement of sets of 25 points in a surface plate of high quality and adjustment
computations according to the method of least squares, using three parameters, the standard error of
unit weight was found to be s 0 = 2 «m. In each point three settings were replicated and the average
was used as measured value, the standard deviation of which is about 1 um.
From the data on precision and accuracy mentioned, tolerances for the standard deviations and
standard errors of unit weight have been established, as a standard before a new operator begins to
measure flatness. For example, the standard error of unit weight, as computed from measurement in
25 points of the surface plate mentioned, should not exceed 2.5 /um (22 degrees of freedom and level
5 per cent).
Reference. 1.34:1. Hallert, B.: Determination of the Flatness of a Surface in Comparison with a Control
Plane. The Photogrammetric Record, Vol. Ill No. 15, 1960.
1.341 Tests of the Glass Grid Wild nr 410
This glass grid was specially ordered. The format is 9" X9" and the spacing is 10 X 10 mm. After a
great number of test measurements in different instruments the standard error of the grid coordinates
was found to be less than 1 in each direction.
The flatness of the grid side of the glass plate has been determined under varying assumptions as to
location of the supporting points. The results are shown in fig. 1.341:1 and 2 in the Appendix. The tolerance
for flatness with respect to the accuracy of the measurements has not been exceeded at any point. These
measurements and computations were made by P. Kaasila, MSE.
1.342 Diapositive Plates
Most restitution work in photogrammetry is made from diapositive plates which can be contact or
projection printed from the negatives. In projection printing, regular errors, especially radial distortion
effects, are frequently corrected. In projection printing flatness defects of the diapositives can seriously
affect the plotting, especially when the observations of image coordinates are made orthogonally.
Therefore attention has been paid to the flatness of diapositive glass plates. It may be remarked that
the flatness of a number of diapositive plates was measured in connection with the International controll
ed experiments within Commission IV of I.S.P. 1958—1960. The report from Sub-Commission IY:4
contains the results of flatness measurements of diapositives with formats 15x15 cm, 18x18 cm and
23 X 23 cm. See reference 1.342:1. In particular, the 23 X 23 cm diapositives showed considerable deform
ations which may cause serious errors in the final results of plotting in instruments with optical or
optical-mechanical projection. The diapositives are contact printed.
In controlled experiments of Working group IV:1 I.S.P. 1960—1964, two sets of contact diapositives
of aerial photographs to be used in the experiments were tested for flatness. The two sets are denoted
1:26, 11:26, 1:28 and 11:28. The results of the test measurements are shown in fig. 1.342:1 through
1.342:10 in the Appendix. In all cases the deformations are based on a plane through three points. The
measurements were made independently in all points and then the residuals were computed with three
parameters, viz., two rotations and one translation. The basic standard error of unit weight of the mea
surement was about 2 ^m and the significance limits of the residuals became d: 6 ¿<m. This means that