539 
also certain checking routines, such as a polygon closure condition, are 
incorporated in the data acquisition software, the digitized data are 
generally of good quality and at the present no need is foreseen to provide 
each data collection station with an editing facility. The editing, instead, 
is carried out as an off-line operation, performed only after the different 
models have been joined. A Tektronix interactive terminal is used for that 
purpose. A disadvantage of the off-line editing is that, in the case of 
serious omissions or errors, the stereo-orthophotos may have to be set‘ up 
again in the Stereocompilers to complete, or correct the data. The 
orientation process on the Stereocompiler, however, is simple, consisting 
merely of aligning the orthophoto and the stereomate on the image carriers and 
remeasuring the ground control image coordinates, which normally can be done 
in about five minutes. 
INITIAL TEST RESULTS 
The tests, which have been performed so far, are intended to demonstrate the 
overall accuracy of the stereo-orthophoto technique including the effect of 
the X-parallax to height ratio used in the stereomate production, the accuracy 
of the process at IGAC to produce the stereo-orthophoto enlargements, the 
instrumental accuracy of the Stereocompilers and the interpretation accuracy 
of the boundary information. 
Photographs of the NRCC Sudbury test area, containing a dense net of 
signalized control points, and photographs of two different Colombian areas 
were used. The accuracy analysis of the Colombian stereo-overlaps was based 
on approximately 75 points which were artificially marked in one of the 
transparencies used for the stereo-orthophoto production. The terrain 
coordinates of these points were determined from stereo-comparator 
measurements and the available ground control points. 
The Sudbury stereomodel was used to test the accuracy of the various technical 
operations at IGAC including the production and enlargement of the stereo- 
orthophotos and the instrumental accuracy of the Stereocompilers. In order to 
obtain the necessary comparative data, the contact size stereo-orthophotos 
were measured on a Zeiss PSK Stereocomparator while 2.5x enlarged stereo- 
orthophoto transparencies were measured on the NRCC Stereocompiler, which is 
of a similar design as the Stereocompilers at IGAC. The values of the 
photogrammetric base (at image scale) and the calibrated principal distance, 
indicated on the GPM orthophoto negatives, were used to convert the X- 
parallaxes measured in the stereo-orthophotos, into heights. In each model 
half of the number of ground control points were used to compute the 
orientation parameters for deriving the terrestríal coordinates and heights 
from the measured image coordinates. The other half served as check points. 
No significant differences were found in the residuals for both groups of 
points and the indicated rms errors represent the combined result for check 
and control points. The calculation of the terrestrial coordínates from the 
stereo-orthophoto model coordínates was based, in all tests, on a linear 
conformal XYZ transformation. It was found that identical rms errors were 
obtained for the stereocomparator measurements of the contact size stereo- 
orthophoto transparencies and the enlarged copies, measured on the 
Stereocompiler. In both cases, the rms errors, expressed at the scale 
1:16 000 of the original images and obtained for position (m, = /m2 - m2) and 
elevation (n,) for the 92 targeted control points in the overlap, were the 
following: 
m. - 66 um 
52 um 
N "Jg