540 
The above values are in agreement with earlier results from the GPM System 
(van Wijk, 1979). The fact that identical results were obtained for the 
stereocomparator and the Stereocompiler measurements indicates that the 
enlargement process, used at IGAC to produce the Stereocompiler 
transparencies, and the instrumental accuracy of the Stereocompiler do not 
appear to affect the accuracy which is obtained by the system. 
Stereo-orthophotos produced from the two Colombian stereo-overlaps were used 
to determine the stereo-orthophoto accuracy for terrain conditions, typically 
encountered in the project. The results obtained from these models may be 
affected by vegetation, which for the Sudbury test area is virtually non- 
existant. Also the elevation differences relative to the flying height for 
one of the models is more representative for Colombian terrain conditions. 
The influence of using X-parallax to height ratios, different from the base to 
height ratio, in the production of stereo-orthophotos, was evaluated for the 
Colombian test models and a Sudbury model, different than the one used in the 
previous test. Using a smaller ratio will result in smaller X-parallaxes in 
the stereo-orthophotos, improving the image quality of the stereomate, but 
affecting the accuracy, as demonstrated by the results. The rms errors, 
obtained for contact transparencies of all test models, as measured on the PSK 
Stereocomparator are indicated in Table I. 
  
  
  
  
  
  
  
  
  
  
test model |relative p, /h image rms error at number 
elevation scale image scale of 
difference 
n, m, points 
Sudbury 2.3%H 0.6455 1:16 000 68 um | 49 yum 95 
(= base/height) 
" 2.32H 0.4317 1:16 000. 1 72 um 1 57. um 108 
" 2.32H 0.2891 1:16 000 85 um | 64 um 96 
Colombian 8.4ZH 0.6254 1:20 000 81 um 76 yum 56 
. (= base/height) 
" 8.472H 0.2811 1:20 000 98 um |112 um 60 
. 2.27H 0.5995 1353 000 44 ym | 44 yum 72 
(= base/height) 
" 2.2%ZH 0.2957 1:53 000 45 um | 83 um 65 
Table I 
rms errors for stereo-orthophotos produced 
with different X-parallax to height ratios 
The results indícate the accuracy limits of the stereo-orthophoto technique. 
They have to be compared with the effect of other error sources, such as 
inaccuracies in identification and measurement of parcel and land use 
boundaries, which may affect the final accuracy. This was evaluated for a 
test with approximately 85 typical boundary points which could be identified 
in the photographs by natural features. The test was performed on stereo- 
orthophoto transparencies, enlarged by a factor 2.5 with respect to the 
original 1:20 000 scale photographs. Two operators identified and digitized 
the boundary points each two times. No significant difference in the results 
was established for the two operators and their combined result indicate 
standard errors of 100 yum in position and 57 um ín elevation, expressed at the 
original image scale 1:20 000, and corresponding with standard errors of 2.0 m 
and l.l m, respectively, in the terrain. The planimetric error in identifying 
and digitizing non-targeted boundaries exceeds the stereo-orthophoto system