: S 8 S 
Number of 0 9 "Stand. dev." 
Instrument Base average max min : 
tests a vis . microns 
microns microns microns 
Autograph A7 Zero 15 5 8 4 1 
in 12 6 7 3 1 
out 6 6 10 i 
Stereoplanigraph C8 zero 18 8 11 5 2 | 
| 
in 5 7 10 5 2 
out 8 7 8 4 1 
  
  
  
  
  
  
  
  
  
Obviously the standard errors of unit weight of the image coordinates are of about the same magni- 
tude for both the instruments. The order of magnitude is about 6 microns. Of course, the errors 
of the grids are also included in this figure. No information has been available concerning the accuracy 
of the grids. The figures in the column for “standard deviation” are computed from the deviations 
of the individual standard errors from their averages. This is done only to give an impression of the 
order of magnitude of the deviations. 
2. THE Y-PARALLAX TESTS 
Introduction 
By measurements of y-parallaxes in sufficiently well oriented photogrammetric models and simple 
computations, valuable information on the general accuracy conditions of the photogrammetric 
procedure can be obtained, see [2]. 
In principle the reconstructed bundles of rays from which the model is created must be free from 
systematic deformations. The irregular deformations of the bundles of rays and the errors of the 
measurements are estimated as the standard error of unit weight of the y-parallax measurements 
according to the method of least squares. This factor is of basic importance in the formula systems 
for the estimation of the accuracy of the final coordinates and elevations and arbitrary functions of 
adjusted model data. 
It is also possible under normal circumstances, however, to determine some types of systematic 
errors, primarily the radial distortion of the reconstructed bundles of rays from y-parallaxes which 
have been measured in suitably located points, see [5]. For obvious reasons, emphasized in [5] and 
[6] this method for the determination of the radial distortion must be regarded as an approximate 
one, which cannot be used for a complete camera calibration but which nevertheless can give very 
valuable information about the geometrical qualities of the model. There are also other factors which 
are of importance for the reliability of the error propagation formulas. First, the completeness of 
the relative and also of the absolute orientation is of great importance. Although the error propaga- 
tion formulas according to [2] are in principle valid only for flat ground and vertical photographs, 
they can also be applied under more general conditions provided that the relative and the absolute 
orientation are sufficiently well performed. This will be empirically demonstrated below. But if the 
orientation procedures mentioned are only approximately performed, as for instance in a mirror 
stereoscope, the elevation differences and the inclinations may have a disastrous effect upon the 
reliability of the error propagation formulas. 
2.1 Measurements of y-parallaxes 
2.11 Position and notation of model points 
For accuracy investigations it is required that the y-parallaxes shall be measured in at least 15 
points, regularly distributed over the model (Fig. 3). Ten redundant y-parallaxes will then be ob- 
tained. If at the same time possible radial distortion of the reconstructed bundles of rays is to be 
6 
de 
cir 
de 
Fc 
pr 
th 
15 
be 
Y- 
tic 
th 
tic 
re: 
re 
ar