ANNER 
UMAX 1200 
mp = £0.15 
tions dy, dy, 
termined dy 
J the sensor 
20 mm. The 
cording with 
d of CCD 
ion 600 dpi 
m long). 
tion equals 
je maximal 
i. It means 
e, the pixel 
ded optical 
med for the 
light. Such 
' photo size 
in only 6096 
h area of a 
scanner a 
e pattern is 
5x5x0, 1mm 
ouped in 16 
| located in 
onsification, 
Imm) were 
ne COREL 
h resolution 
e pattern is 
The pattern made on film was then measured on precise 
stereocomparator Stecometer with accuracy my = my = 
+3um. The results of this measurement were treated as a 
ideal set (pattern) against results achieved by measuring 
digital form of the pattern. 
The pattern was scanned with resolution 1200 dpi (pixel 
size 0,02x0,02 mm). Such resolution was admitted as 
optimal for digital photogrammetry measurement using 
Video Stereo Digitizer AGH (VSD) [Jachimski 1995]. The 
pattern of size 15x22 cm scanned with above mentioned 
resolution has a capacity above 70 Mb. 
Scanning of such an image takes approximately 20 
minutes. The method of the scanner accuracy testing 
described in [Boron,1994] and [Boron,1995] assumed 
scanning of the pattern in two sessions, one by one, 
without switching off the scanner. Such a procedure 
allowed for smaller scanner errors but unfortunately 
caused deformation of pattern (marked on the 
transparent film) under high temperature of the scanner 
being heated. 
To eliminate that disadvantageous phenomenon there 
were applied at least one hour brakes between pattern 
scanning sessions, to cool the scanner completely. The 
pattern was so placed on the scanner plate that the 352 
pattern crosses were equally covering the scanner 
working surface, and the lines of crosses were parallel 
to respectively X and Y scanner axis. In such position 
pattern film was fixed with the tape to the scanner plate. 
The steady pattern position allowed for the analysis of 
repeatability of scanner errors. 
During the testing program 22 pattern images were 
scanned, and they were named successively from A to 
Z. During 4 days there were registered: in the first day 
A...| images, second day J to N, in the third day O...S, 
and in the fourth day T to Z. 
The received digital images of the pattern were to be 
measured to register coordinates of crosses (some 8 
thousand crosses) in the digital image coordinate system 
(pixels matrix). For that purpose a devoted program for 
automatic crosses position recognition was prepared by 
W.Trocha (described in [Trocha,1995]. As the above 
program requires the approximate crosses position as 
the entrance data, there was performed with the use of 
VSD a hand survey of 4 corner crosses on each digital 
image. A next step it was transformation of pattern 
coordinates (surveyed with the use of Stecometer) to 
the digital image coordinates (pixel system) using the 4 
corner points as a pass-points for the bilinear transfor- 
mation. 
For each digital image of the pattern there were surveyed 
352 crosses (22 rows, 16 crosses each) which created 
an entrance file for the program of automatic crosses 
position recognition. 
3.THE TRUE ACCURACY OF THE UMAX 1200 SE 
SCANNER 
To determine to what accuracy level the digital images of 
the scanned pattern reflect the true geometry of the 
pattern the Helmert transformation (similarity transforma- 
tion) for the full set of the pattern coordinates (in mm) 
and full set of the cross coordinates in image system (in 
pixels) was used. The results of these transformation are 
given in the Table 1 while an example of geometrical 
deformations of an one digital image against the ideal 
21 
pattern is shown on Fig.2. As can be seen in Table 1, 
fitting errors of points for both sets for particular images 
are similar and vary from mp7 +4,8 pixels to mp= +5,3 
pixels. 
Table 1. The results of the Helmert transformation 
  
Helmerttransformation — 
*(352 reference points) ^ 
my mp (dx)max (dy)max 
(1200 dpi) | pixels | pixels | pixels || pixels | pixels 
9. /2 3.71 5.25 11.1 10.2 
3.67 3.36 4.98 11.0 9.0 
3.67 3.38 4.99 11:5 8.7 
3.68 3.26 4.92 11.3 8.2 
3.55 3.24 4.81 10.7 8.5 
3.50 3.22 4.76 10.7 8.4 
3.55 3.18 4.76 11.0 8.0 
3.56 | 3.23 | 4.81 || 10.9 | 8.4 
3.57 3.24 4.82 10.7 8.5 
3.67 3.52 5.09 10.7 9.4 
: 10.6 8.5 
3.56 3.54 503 10.4 9.7 
3.72 3.41 5.04 10.7 9.1 
3.67 3.33 4.95 10.7 8.8 
3.80 3.59 5.22 11.2 9.6 
3.64 3.34 4.94 10.9 8.7 
3.60 3.22 4.83 11.0 8.1 
3.61 3.22 4.84 11.0 8.1 
3.77 3.44 5.10 11.1 8.9 
3.68 3.30 4.94 11.0 8.4 
3.65 3.26 4.89 11:1 8.1 
3.68 3.29 4.94 11.2 8.5 
[m]/n 3.64 3.34 | 4.94 
Image 
number 
  
   
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
N|ÍSIc|ejo|-3Tololsi3i-—iIx[i-|-|-ko|-/|o |o.]o |o |o 
w 
Cc 
Oo 
C 
w 
O 
A 
œ 
i 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
SCALE 
Mx = 3.67 
My = 3.52 
Mp = 5.09 
  
Fig.2. Deformations of the image J after the Helmert's 
transformation 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996