T mA 
  
——R 
  
  
  
  
  
  
  
  
   
    
    
   
  
  
PEER a 
Ts 
  
  
ipti SERERE ce tee ee 
i 
À 
j 
  
  
  
    
abd uc 
   
— HÀ 
  
It can be seen that the procedure is almost identical with the pr 
: : : : OCedure of , 
triangulation based on use of statoscope data, with the difference that hy DET 
consecutive models not the elevation carry-over points but profile points a» T 
€ Used, 
Adjustment of z, x and y. 
Adjustment of the results must be started with the adjustment of 
recorded z-values of the points must first be corrected by the discrep 
on the elevation carry-over points. Because the elevations of profile 
for the scaling, identical points in two consecutive models will have di 
It is obvious that the elevations from the previous mode 
2 eoordinats p, 
ancies in deri 
Points Wen W 
fferent deni 
; L have to be kept 4 {he 
elevations and therefore the z-readings in every model have to be corrected rm 
It is better in this regard if the corrections are based on z-reading diserepandeg go 
elevation carry-over points chosen close to the principal point of last photograph in SN 
pair. Then, the differences between the corrected ''bridged" elevations of ler] n 
points and their previously ms 
tabulated ‘‘profile” elevations 
are obtained. If these differen- 
ces are plotted against x- 
coordinates of the points, a sort 
of polygon is obtained. 
If profile elevations were 
free of errors, the situation 
would be very simple. We 
would then have a large number 
of true elevation points which 
would determine directly the 
z-corrections for each model. 
Unfortunately, as the ex- 
perience with radar profile 
technique proves, profiles show 
very high general accuracy but 
single points are affected by 
accidental errors due to numer- 
  
  
  
ous physical phenomena in- Fig. 2. 
herent to this technique. 
Exactly the opposite can be said about aerial triangulation. Except for possible ltd 
points, the “local” accuracy of aerial triangulation is very high, whereas general au 
is limited because of unfavourable accumulation of various errors. Consequently thet 
rection curve must be a smooth curve and not a polygon, as it would be if profile elev 
affected by accidental error were accepted. Strict analytical determination of this sui 
curve is quite a difficult problem. However, a very simple graphical of numerical sii 
is possible, applying a smoothing procedure as used in treatment of experimental 0 
For instance, by replacing polygon points by the points which are centres of gni 
certain number of consecutive polygon points, such a smooth curve can be obtu 
Repeating the smoothing proces or using a larger number of polygon points for ouf 
tation of centres of gravity, higher degree of smoothness will be obtained. | 
This method, which can be applied very successfully to various photogranté” 
procedures, has been used for the z-adjustment in our experimental aerial triang 2 
The curve of the twofold repeated smoothing process using each time centres of gii 
of the three consecutive points, was accepted as the correction curve. d 
The theory of smoothing provides means to determine the necessa deg! 
: STO vation der 
smoothing. This is based on a comparison between the mean erro! of observa 
  
  
   
   
   
  
  
   
   
   
   
  
  
  
   
  
  
   
   
    
   
indications 
ELEVATIONS 
FOR 
CURVE 
Smoothing of APR- 
© Correction according io APR 
— Correction Curve 
Established by 
CORRECTION