dan a pet 
n PC 
UL. aß 
Fig. 4: Two tablets 
  
  
  
  
  
  
  
  
The observation unit is a mirror stereoscope mounted 
on a track attachment and the two cursors should be 
assembled to a device similar to a parallax bar. 
The plotting software runs from AUTOCAD through an 
AUTOLISP program allowing, in this way, 
simultaneous plotting. 
Before running the plotting software, an AUTOLISP 
program whose name is RESTI, inner, relative and 
absolute orientation are carried out running the 
program ORIENTA written in QUICK BASIC 45. After 
measuring the four fiducial marks of each photograph 
for inner orientation, relative and absolute orientations 
are performed using practically as many points as 
wished. Points can be measured in arbitrary order and 
positions. For control points the identification number 
will be required, and this number will be the link with 
the ground coordinates of the point which are in a 
preexistent file. The computed parameters of all 
orientations will be saved in a file, which will be used 
afterwards by RESTI. 
As AUTOCAD recognizes only one tablet, RESTI 
overcomes this difficulty by calling a routine in 
QBASIC, which can handle two tablets without 
problems. Using this routine a cartographic feature is 
measured, point by point, as in an Image Space 
Plotter, generating a file. When the measurement of 
the feature finishes, the routine is also finished, and 
the control comes back to the AUTOLISP program, 
which will then read the file containing the feature 
points and plot a polygon describing it. The edges of 
this polygon can be smoothed afterwards with the 
AUTOCAD command SPLINE, if wished. Thus, feature 
by feature, the map can be completed. Being an 
AUTOCAD file, the map can be read without problems 
from several GIS. The hardware arrangement, which 
was implemented for making some proofs, consists of 
a Sokkisha mirror stereoscope mounted on a track 
attachment, with the two cursors placed ‘on the 
parallax bar. 
There is another version of the program RESTI for 
monoscopic observation written entirely in AUTOLISP 
that can be used for procedures explained in 2.1 and 
2.2, which use only one tablet. As it was said before, 
homologue points should be measured with the same 
cursor in both photographs, one after the other. The 
same program could also run for sterescopic 
observation if the arrangement of fig. 3 is used, a 
possibility which is also considered in our project. 
To build an efficient Image Space Plotter, according 
with these ideas, there are still other problems to be 
solved. One of them is the cross hair of the cursors, 
which is unfavorable for stereoscopic observation. A 
small circular dot would be much better. Another 
problem is how to arrange the tablet, the stereoscope 
and the track attachment in order to meet elementary 
ergonomic requirements. We are trying now to solve 
these problems and expect that the resulting plotter 
could be an interesting tool for mapping. 
3. CONSIDERATIONS ABOUT THE ACCURACY 
The accuracy will depend essentially on that of the 
digitizing tablet. Warner and Carson ( 1991) have 
tested the GTO Digi Pad tablet of an area of 45 x 63 
cm (cost : less than $1000). The tests were carried 
measuring an accurate grid plate, and the results 
confirmed what was claimed by the manufacturer : a 
resolution of 25 um and an accuracy of 87 um. The 
quoted authors report also that removing systematic 
deformations, the 87 um were reduced significantly, 
remaining only 58 um. 
In our Institute we have tested a Summa Sketch Ill 
tablet of an area of 30 x 30 cm and a cost of about 
$400. The manufacturer claims for this tablet a 
resolution of 25 um and an accuracy of 250 um. Our 
test showed that the deformations were strongly 
systematic. After measuring an accurate grid an 
orthogonal transformation between measured and grid 
values was performed. Figure 5 shows the pattern of 
the resulting deformations ( coincidence in points 11 
and 19 was forced ). It is clear that the deformations 
are strongly systematic. It is also clear than after 
removal of this systematic effect, accuracies similar to 
those obtained by Carson (better than 100 uim) can be 
obtained. 
11. 582^ 543; 14 15 16: 647 : 185 (119 
21 Q 23 24 25 26 27 28 29 
61 62:63; 64V 100525:66 5 87:68:60 
— 1mm 
Fig. 5: Deformations in a low cost tablet 
Distance between grid points, 2 cm 
At present, we are testing a Micro Grid Il from 
Summagraphic, which is somewhat expensive ( about 
396 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996