Jevi- 
osed 
used 
II be 
reo- 
uted 
mine 
test 
cali- 
natic 
id y 
eica- 
t by 
sys- 
ving 
f the 
with 
by a 
three 
nt to 
tion. 
the 
the 
ffect 
case 
  
When processing the measurements it has been 
assumed that the systems can be corrected for the 
effects which are described above. 
To correct for the systematic deviations a trans- 
formation has been applied on the observations 
with as parameters two shifts, two scale factors 
and one rotation. Patterns of systematic deviations 
are given in figure 4. 
Systems Zoom 
1 10x 
20x 
40x 
5x 64 
10x 64 
12 
10 
  
With different zoom settings the line width is deter- 
mined at the circle position by setting the measur- 
ing mark at two positions of the line (see Fig. 5). 
Through the knurled character of the lines (raster 
screen) a measuring mark as large as possible has 
been choosen. For the computation of the line 
width it was necessary to determine the diameter of 
the measuring mark too. 
22 14 
16 10 75 
Table 5. Test B. Standard deviations in microns at photoscale before 
0, , O,' = after correction; n = 
Table 5 shows that the standard deviations after 
the correction for systematic errors become con- 
siderably smaller. When using a 40 times magnifi- 
cation the smallest residual errors are present in 
system 1 (0, = 10 4, 0, = 10 yu). For a zoom factor 
of 10 times the smallest standard deviations are 
present in system 3 (0, = 16 yu, o, — 10 y). 
The conclusion is that for the detection of data 
base errors all the used systems should be calli- 
brated. 
3.4. Test C. Determination of the width of the 
superimposed lines. 
The factors which influence the width of the super- 
imposed lines are the illumination, the line direction 
and the zoomfactor. 
When the superimposed line is strongly illuminated 
with reference to the photo then the line will be 
thicker than in the case of a poor illumination. 
When the line is represented on a raster screen, 
then horizontal and vertical lines will have a width 
of 1 pixel. In other directions this can increase to 
1,5 pixel. 
Considering the above mentioned factors a line 
pattern as shown in figure 5 has been used for 
testing. 
and after systematic error correction. o,, 0, = before correction; 
number of measured points. 
567 
Because the lines on the screen move in steps of 1 
pixel measuring problems occured. To get a good 
measuring accuracy every point has been measured 
four times. 
  
  
  
Figure 5. Superimposed line pattern for line width 
measurements. . = measured point.