grid intersections were not a problem, except in only a couple of 
instances. Nine regularly spaced camera stations in classical 
configuration approximately 2 metres from the digitiser were used to 
acquire full coverage over the grid (refer to Diagram A). The camera 
was mounted on a stand, and coverage confirmed before exposing 
the film. The film was processed through a public commercial 
laboratory. 
ue ut 
  
  
  
Digitiser 
  
  
  
  
O-— > 
Camera station 
and direction 
Diagram A: Camera-Digitiser Geometry 
5.3 Image Coordinates 
Image coordinates of the grid intersections were measured on a Zeiss 
Planicomp C100 analytical plotter using the black and white negatives. 
Lens distortion, but not film unflatness, was allowed for during the 
measuring procedure. 
5.4 Bundle Adjustment 
Grid intersection coordinates were determined by bundle technique 
using the image measurements of the nine photographs. The 
adjustment was constrained by holding the central grid point fxed 
together with one other point for "azimuth". The digitising table was 
assumed to be reasonably flat and a number of height points were 
loosely held. Scale control was provided through measurement on 
three one-metre scales. 
5.5 Results 
The result of the bundle adjustment was a set of coordinates for the 
grid intersections. The one sigma level of accuracy for XY (in the 
plane of the digitiser) was about 0.1mm while for Z (perpendicular to 
218 
the plane of the digitiser) was about 0.2mm. Sigma nought for image 
measurements was 5 microns, which was considered satisfactory for 
this project. 
6. ANALYSIS OF COORDINATES 
The digitised coordinates (Dig 7 and Dig 10) were shifted and rotated, 
but not scaled, to fit the nominal values (even 100mm) of the 
reference grid. The X and Y differences between the transformed 
coordinates and the nominal grid values were determined and 0.1mm 
contours of the separate X and Y differences plotted. These results 
are shown in Figures 1 to 4. 
As can be seen from the plots, the two manual digitiser results are 
similar to each other, but differ from the reference grid by a 
considerable amount; some half a milimetre in both X and Y 
directions. At this stage it appeared there was a simple uniform scale 
error in the digitiser. 
Next the photogrammetric coordinates of the intersection points were 
compared with the nominal values of the reference grid. 0.1mm 
contours of the X and Y differences are shown in Figures 5 and 6. 
The differences are similar to the manual digitiser results, some half 
a millimetre in both X and Y directions. 
Contours of the X and Y differences between the manual digitiser 
coordinates and the photogrammetric coordinates are shown in 
Figures 9 to 12. As could be expected, the results are in agreement 
with the differences, and more or less consistent with the 
photogrammetric accuracy of 0.1mm. 
The scale for the photogrammetric coordinates was derived 
independently from metal scales and not from the reference grid. The 
conclusion to be drawn is that the reference grid contains a scale 
error. 
Finally, the nominal grid coordinates were conformally transformed to 
fit the photogrammetric coordinates. The scale of the grid was 
calculated as 1:0.9993. Contours of the X and Y differences are 
shown in Figures 7 and 8. 
At this stage the project appeared to change from one of digitiser 
calibration to one of reference grid calibration. 
7. CONCLUSIONS AND RECOMMENDATIONS 
The project was successful in that we verified to 0.1 millimetre, the 
measuring accuracy of our digitising system. 
However, the project highlighted a problem with our "standard" 
reference grid. Had we adopted the reference grid as correct then 
we would have mistakenly thought the digitiser to be in error. 
A solution to the calibration problem for a large table digitiser would 
be for the manufacturer to mark a grid on the table itself in addition 
to boundary marks, similar to the grid etched on a photo carrier of an 
analytical plotter. The calibration of the permanent grid could then be 
done efficiently and conveniently by close-range photogrammetry. 
8. ACKNOWLEDGMENTS 
The authors wish to thank Mr Andrew O’Dempsey of the Redland 
Shire Council, Queensland, for his input while he was a research 
associate with the Queensland University of Technology. 
9. REFERENCES 
Burrough, P.A., 1986. Principles of Geographical Information Systems 
for Land Resources Assessment. Clarendon, Oxford. 
Rollin, J.R., 1986. A Method of Assessing the Accuracy of 
Cartographic Digitising Tables. The Cartographic Journal Vol.23 No.2. 
Slama, C.C. (Ed), 1980. Manual of Photogrammetry 4th Edition. 
American Society of Photogrammetry, Falls Church, USA.