o 
1g 
RC5a 11.5/18. This can to some extent be explained by possible affine deformations of the models due 
to uncorrected shrinkage of the film negatives, see above under point 2.42. Owing to these circum- 
stances and to the large elevation differences, the y- and z-coordinates in particular can easily be 
affected by systematic errors. Moreover, according to Sub-commission IV:1, the coordinate trans- 
formation from the models to the ground was made exclusively with one scale factor. 
4. The differences between the results of Sub-commissions IV:1 and IV:4 are all very well in- 
cluded within the confidence intervals for the confidence level 1 per cent and also within the con- 
fidence intervals for the 5 per cent level, except the y and z of models from the camera RC5a 11.5/18 
which exceed the upper limits by a few millimetres. 
5. The hypothesis that measured y-parallaxes can be used for a prediction of the accuracy to be 
expected in the final coordinates (under certain well-defined conditions, in particular concerning the 
fundamental operations Nos. 1 and 2, as clearly stated above) is therefore practically proved once 
more, this time under unfavourable external conditions (large elevation differences and large in- 
clinations of the photographs). Consequently, measurements of residual y-parallaxes in oriented models 
are of great value for the estimation of photogrammetric accuracy in general. 
Recommendations 
In accordance with the results of the investigations above, the following can be recommended. 
1. It is desirable that the fundamental operations of photogrammetry (also the separate components) 
be investigated separately and in detail, in particular for the determination of regular errors and 
estimation of the irregular errors. Grid tests of aerial cameras over flat and rugged ground would be 
particularly valuable. International cooperation is strongly recommended. 
The atmospheric refraction may be one of the limiting factors, especially over rugged ground as 
at Monti di Revóira. The better the regular errors of the fundamental operations are determined 
and the better they can be corrected for, the better will the laws of the propagation of the irregular 
errors of the fundamental operations agree with practice. 
2. Tolerances for the errors of the fundamental operations should be constructed after a sufficient 
number of tests of the operations. The tolerances must be founded upon statistical laws for error 
propagation in which the expressions for the basic accuracy of image coordinates and coordinate 
differences (parallaxes) are inserted. A certain confidence level should be established. 
ACKNOWLEDGEMENTS 
We wish to extend our sincerest thanks to those participants in the controlled tests who have 
contributed to the work also of Sub-commission IV:4. 
Furthermore we are most grateful to the following persons for valuable help: Messrs. B. Blomstedt, 
B. Dryselius, K. E. Eriksson, Dr. C.-G. Esséen, Mrs. G. Fagerholm, Messrs. B. Gustafsson, N. Gyldén, 
R. Helming, B. Malmsten, T. Thedéen, A. Thunberg, and Mrs. E. Wernholm. 
20