raphs, 
oots 
of 
to 
y 
not 
s of 
ation 
rtance 
n 
ng 
are 
sure- 
|. Physical influences, which deform the central perspective bundles and 
therefore devaluate the condition of collinearity (compare Chapter III), are 
lens distortion and refraction. The lens distortion can be determined along 
with the geometrical parameters (See Chapter VI). Star or collimator pho- 
tography, in connection with the analytical treatment of an individual photo- 
graph, provides a practically unbiased means of determining the calibration 
of such a camera ; that is, its interior orientation and the specific lens 
distortion. The simultaneously obtained elements of exterior orientation allow 
the calibration of phototheodolites or the establishment of the mutual relation 
between several cameras a8, e.g., is necessary in connection with the use of a 
sun camera. (See Chapter IV-C) 
Based on practical experience, it can be generally stated that the analyti- 
c&l treatment appears to be an excellent means for analyzing objectively the 
components of the photogrammetric measuring method. 
If refraction can not be eliminated by a suitable arrangement of the 
measuring set-up, its influence must be eliminated for each individual ray, 
making use of specific meteorological measurements or assuming & certain model 
of the atmosphere. Information concerning the computation of the corresponding 
corrections and their consideration in the analytical reduction method is given 
in Chapter VII. 
Ignoring the economical side of the problem, in precision photogrammetry 
the requirements for flatness and dimensional stability of the emulsion carrier 
can be satisfied by the use of precision ground glass plates. The unavoidable 
irregular shrinkage of the emulsion, together with the measuring errors on the 
comparator cause the plate coordinate measurements to be affected by residual 
errors, which have a distribution similar to a normal distribution, thus justi- 
fying the effort of a rigorous least squares treatment. Such a computing 
technique will also prove to be practical in cases where systematic errors are 
present. Not only does a least squares solution in any case produce the most 
likely result, but additionally obtained qualitative and quantitative infor- 
mation allows one to recognize and isolate systematic errors. In addition, 
the least squares treatment, as described in this report, is simple, due to 
TT