>: R, height, 
)ands 
  
S X, 
e bands 
5.1 
5.2 
= AO 
The most recommendable method for practical purposes remains 
however to correct systematic image errors as far as possible 
before the process of triangulation commences. This is by far 
the simplest method to eliminate a large part of the systematic 
errors. With the modern process tools available at present and 
with proper handling, the remaining systematic errors should 
in most cases not be dangerous for conventional mapping projects. 
These components are only dangerous in special applications 
and their effects can then be eliminated by one of the above 
- in general elaborate - methods. . 
TOWARDS A MORE COMPLETE THEORY OF IMAGE ERRORS 
  
At the very moment we defined systematic image errors, 
we realized the incompleteness of our definition. The definition 
of the random image errors is to the same degree known to be 
incomplete, As a matter of fact, so far no description of the 
image errors is known, which could claim a reasonable degree. 
of completeness. Let us now try to give some indications as to 
the theory on which such a description could be based in future. 
It is generally known that the image coordinates of points 
in one plate j, which are strongly correlated, are. also 
correlated with the point-locations in the adjacent plates. 
This correlation is reasonable, since the adjacent images 
were taken within a short interval, in which the state of the 
atmosphere and of the camera system was only changed slightly, 
and the treatment of those images during the photographic 
processing was probably quite similar (cf.Ziemann 1971). 
We assume that those factors which influence the image locations 
change in a random manner in time and space, and we now aim 
at a mathematical description of these factors. For this 
purpose the concept of a random function of one variable is 
introducea (for an authorative treatment of the subject cf. 
Felier 1966): 
We shall define a random function as a function X(t) of a real 
variable t, the value of which for any value of t is a random 
variable. The argument t will be considered as a non-random