International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004 
  
3.4 Test of calibration results 
Obtained calibration parameters of scanner are tested by 
resampling of grid plate using these parameters in exactly the 
same way as it would be done for scanned images. Providing 
that previous conclusions were correct, only random parts of 
overall errors should remain on rectified grid plate. 
Transformation results after rectification are given in following 
table. 
  
Affine Collocation 
My Mx My Mx 
(um) (um) (um) (um) 
platel 10.48 11.51 3.25 3.67 
plate2 11.10 12.77 3.12 2.37 
Maverace | 10.79 12.14 3.18 3.27 
Mmax 11.10 12.77 3.25 3.67 
Mmin 10.48 11.91 3-11 2.87 
Range 0.62 1.26 0.13 0.80 
Table 10. RMSE after plate rectification 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
It is visible from table 10 that after rectification of plate 
according to obtained calibration parameters remaining part of 
mean positional error is 11 um to 12 um. This means that 
remaining errors for images scanned and in this way rectified 
would be the same. 
3.5 Practical experience 
In addition to experimental research based on described 
technology, about 3000 aerial images were processed for the 
requirements of digital orthophoto production in Serbia during 
the last 3 years. 
Interior orientation and aerial triangulation, procedures which 
also provides good check of geometrical quality of scanned 
images, has confirmed conclusions stated so far in the paper. In 
the process of automatic interior orientation RMSE of the affine 
transformation is around 0.9 pixels, with maximum error of 1.2 
pixels. Aerial triangulations with bundle adjustment for several 
projects with 100-400 images resulted in RMSE about 1 pixel, 
for GCPs and tie points. 
Since mentioned scanner has no system for film alignment, it is 
necessary to press film negative by glass plate during scanning 
process. Also, the there are scanning problems related to 
appearance of Newton rings. This problem can be removed by 
using AntiNewton glass plates or by applying some other 
known method. 
4. CONCLUSION 
The high price of professional photogrammetric scanners at one 
side, and the requirements of market for various products of 
digital photogrammetry at the other side, calls for very careful 
approach when one is trying to start autonomous digital 
photogrammetric production. 
The major drawback with desktop scanners when used for 
digital photogrammetry is geometric accuracy. Nevertheless, 
testing and calibration of EPSON Expression 1640XL showed 
that with additional calibration this type of scanner can be used 
in photogrammetry. 
By multiple scanning of glass grid plate, and by using linear 
prediction by least squares method the following conclusions 
can be made: 
9 ‘there are large geometrical scanning errors 
(RMSE=130 pm), so this scanner can not be used 
without proper calibration 
e after removal of systematic part scanning RMSE is 
reduced to 4.2 um and 8.5 um 
e systematic errors are very stable, regardless of 
duration of scanning procedure and changes in 
geometrical and radiometric resolution. It has been 
estimated that variations of RMSE after affine 
transformation is 3.2 jum (Y axis) and 5.1 um (X axis). 
e it was proved that systematic errors in 9 distinctive 
points are very stabile, regardless ofo duration of 
scanning procedure and spatial and radiometric 
resolution changes 
e it was estimated that remaining RMSE after image 
calibration/rectification by using presented procedure 
is about 12 um 
Three years of using described procedures and scanner EPSON 
Expression 1640XL, confirmed all the conclusions stated 
above. 
However, drawback for using desktop scanners in digital 
photogrammetry lies in the fact that they are still not suitable for 
high precision tasks, because estimated accuracy limit of 12 um 
is insufficient for many photogrammetric procedures. Also, it is 
very difficult, if not impossible, to automate the whole scanning 
procedure by using desktop scanner. It makes scanning rather 
slow compared to professional photogrammetric scanners. 
Although desktop scanners cannot be recommended for large 
and high precision projects in digital photogrammetry, they 
could be used for smaller projects and where highest accuracy is 
not required. With additional calibration these scanners could 
be optimal solution considering cost-effectiveness. 
5. REFERENCES 
References from Books: 
Moritz H., 1978. Aproximation methods in Geodesy. Hebert 
Wichmann Verlag Karlsruhe, pp. 44-87. 
References from Other Literature: 
Baltsavias E., 1996. DeskTop Publishing Scanner, OEEPE- 
Workshop on Application of Digital Photogrammetric 
Workstation, Lausanne, pp. 75-98. 
Jacobsen K., 1996. Experiences with Rastermaster RMI. 
OEEPE-Workshop on Application of Digital Photogrammetric 
Workstation, Lausanne, pp. 99-109. 
   
    
  
   
   
    
     
   
    
      
    
   
   
   
  
      
    
   
   
    
    
     
  
    
   
   
  
   
  
  
     
  
   
   
    
  
   
   
  
  
   
     
   
   
   
   
   
    
    
  
    
    
  
KEY WO 
ABSTRA« 
Little rese: 
photogram 
and graphi 
variable ar 
practical e 
A precisio 
variability 
scanners. 
production 
the errors 1 
During the 
though lar; 
scanned ir 
obtained fi 
1 INT 
With the 
attention | 
capture. F 
of airborne 
overwheln 
conventior 
are collec 
converted 
for use in | 
The choi 
introduced 
photogram 
results obi 
Model (D 
imagery tc 
Li Jus 
A distinct 
(PG) scan 
the mark 
standards 
capability