on type A, 
200 um 
els. 
- 500 um e 
o 
5 - 
It appeared however, that the deformation reconstructed in 
this way is much smalier than the original model deformation. 
This can also be proved within the theory of least squares 
adjustment. Reasonably good results are only obtained with 
dense control in the strip or block. 
The most appropriate method is of course to examine the 
theoretical and the estimated factual n-dimensional distribution 
function of the residuals after least squares adjustment. Using 
the differences in size and shape between these distribution 
functions it can be checked whether the original assumptions 
regarding the error properties of the image or the model 
coordinates hold. Both methods described above are partial 
methods within this general class of tests. 
The elimination of the effects: 
  
Beside the problem of detection there is the even more important 
problem of compensation. The ultimate aim in practice is of 
course not only the location, but also the elimination of the 
effects of systematic image errors. These two problems of 
detection and elimination are strongly interrelated, as an 
elimination pf error effects seems only justified if significant 
effects are detected. 
A,very general and elegant method of computation is to extend 
the mathematical model of the projective relationships and to 
include also terms for systematic image errors (cf. Schmidt 
1971, Masson d'Autume 1966, Cenan 1970). 
These extended projective relationships are then as follows: 
a, (X-XS)+a,(Y-YS)+az(Z-25) 
  
  
  
€ > a i J 
V_+X-X, = C $+ £0 (2) (g~Y-) 
x 0 a (X-XS)+ag(Y-YS)+ag(Z-ZS) ij 1j' ^9 0 
a, (X-XS)+a (Y-YS)+a, (Z-Z5) i 
V1-39 = oA 2 + 1554 4(X-X9) (y-yg)? 
a, (X-XS)*ag( Y- YS) «ag( 2-28) 
e / A iy 
conventional formulae extension for 
systematic image 
errors 
(X, Y, 2) coordinates of a terrain point in an arbitrary 
rectangular terrain system