olution is, in 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
In [Boroñ, 1995] the sinusoidality of d,(Y) discrepancies 
  
To connect the surface of l-order correction to the 
currently scanned digital image it is enough to register 
   
     
  
Fig.3 can be was presented for UMAX 1200. That phenomena can be 
in each line a result of non proper synchronisation of the sensor-line ^ and survey one pass-point targeted on the scanner plate. 
arious lines. position with the registration frequency. Comparison of To actually determine the ll-order corrections the perpen- 
sinusoidal discrepancies function for various digital dicular to sensors row a line of targeted points is needed. 
images [Boror, 1995] show that they are similar, but not — The ll-order corrections are being calculated individually 
yrmation identical (see Tab.3: standard deviation of the average for selected rows of pixels (Y=const) and than interpo- 
error my is great, and equals to over 20%). lated for the other rows. The value of the ll-order correc- 
on | tion is constant for all pixels located in one row Y=const. 
Vo 
x | Odin 5. THE METHOD OF GEOMETRICAL CORRECTION To get average correction function dy, dy the residual 
s | pixels OF DIGITAL IMAGES errors after Helmert transformation were analysed for 
e 3.6 the 14 pattern images (A...N). By averaging 14 values of 
: The correction is executed in two stages. In the first stage —— discrepancies for each of 352 surveyed points of reseau 
30 are determined l-order corrections dy, dy, being steady for ^ pattern the table of average correction function values 
3.2 each point of the scanner plate and determined during — was created. Using SURFER program a graphical repre- 
2.4 pre-calibration process. Next are determi-ned dy correc- sentation of average correction values was created for dy 
2.7 tions of Il-order, being variable for each scanner run, but (Fig. 4) and for dy (Fig. 5). The image of the dy functions 
2.7 on certain digital image being constant along the sensor — shows clearly the sinusoida-like shape of that function 
2.4 line and changeable for different sensor line positions. surface. 
27 
2.6 J 
32 els 
2.8 gX (in p 
3.5 
27 
2.4 3 
3.2 0 
27 3 
2.9 ô 
23 3 
zl. 
2.4 
24 AST 
24 x SS 
5 Aem 
ANT 
  
bilinear 
  
   
Fig.4. Surface of d, errors after the Helmert transformation (images A...N) 
  
Fig.5. Surface of d, errors after the Helmert transformation (images A...N) 
23 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996