'ame 
ry. Tt 
| can 
input 
ation 
ware 
fic to 
and 
D/33 
The 
code 
| 18" 
ound 
Is, as 
etest 
with 
50 to 
ising 
d by 
(1) 
'ARS 
RC-2 
'ARS 
self- 
, this 
than 
‘ARS 
    
simulation software, including both the video-based 
and film-based measurement networks. 
3.1 TEST NETWORK 1 
The first test discussed took place in 1991 and was 
presented in Winnipeg for Commission V of ISPRS 
(Gustafson, 1991). It involves the video measurement 
of a test field and the accuracies achieved from each of 
two target sizes. The test field used is a 7mm thick 
aluminum sheet painted flat black. The targets used in 
this test were 6mm and 11mm diameter circular retro- 
reflective targets. The targets comprise a basic grid 
pattern and additional targets were placed on shims to 
offset points from the surface 10 to 25 mm. To aid in the 
capturing of images, the test field was placed horizon- 
tally on a turntable. Figure 2 illustrates the basic 
  
& eO 
  
  
Figure 2 
Test field and camera geometry 
  
  
  
geometry used in this video test. Note a single camera 
was used and that the test field was rotated to obtain the 
geometry shown. Two rolls were taken at each position 
for a total of 8 stations. 
Computer simulations were made to make accuracy 
predictions based on two levels of image measurement 
precision, the equivalent of 1/20th pixel and 1/50th 
pixel respectively. Resulting accuracy expectations 
were 1 part in 30000 to 1 part in 80000 over the major 
dimension of the area to be measured (1.45M). 
  
RMS Closures of Triangulation 
  
XY Residuals 
: Ratio to Pixel Size 
(microns) 
  
6mm 11mm 6mm 11mm 
  
0.34 0.23 1/20 1/30 
  
  
  
  
  
  
Table 1 
    
Results are reported for both the 6mm and 11mm 
targets. The first data reported is the summary for the 
closures of triangulation for the bundle adjustments, an 
of indication internal precision (see Table 1). 
The second data reported is the summary of accuracy 
results. Results in this case are comprised of the residu- 
als resulting from a three-dimensional rigid-body 
transformation of the video results into the film results 
(see Table 2). The film measurement yielded accuracies 
on the order of 1 part in 300000 (about 0.005 mm in each 
axis) and had an RMS closure of sigma, -0.59um. 
  
Accuracy Summary - Videk vs. CRC-2 
  
6mm Targets 11mm Targets 
  
vX vY vZ vXYZ vX vY vZ vXYZ 
  
0.038 | 0.041 | 0.030 | 0.036 | 0.025 | 0.028 | 0.025 | 0.025 
  
  
  
  
  
  
  
  
  
  
Table 2 
A couple of points of interest should be discussed. To 
begin with, closures of triangulation are improved sig- 
nificantly with the larger target size. As mentioned 
previously, target images are measuredusing a weighted 
centroid technique and the more pixels involved in the 
computation, the better the centroid can be determined. 
The images for the 6mm targets range from 40 to 60 
pixels in area while the 11mm targets range up to 130 
pixels. Also observable is the better accuracy achieved 
with the larger targets. Accuracies of nearly 1 part in 
60000 were obtained with the 11mm targets while 1 part 
in 40000 was the value for the 6mm targets. 
3.2 ANTENNA MEASUREMENT 
A test was carried out in early 1992 to simulate a portion 
of a full scale antenna survey. The final potential 
application of the video technology is as the measure- 
ment portion of a closed-loop shape adjustment system 
fora34mreflector. Thesuggested scenario would place 
several video cameras around the edge of the reflector 
in a more or less permanent arrangement. As many as 
24 cameras would be attached at even intervals around 
the edge looking across at the other side of the reflector. 
Retroreflective targets would been seen in at least 3 
images each, with some in many more than that. Com- 
puter simulations were made to initially establish the 
viability of such a system. In the final system, accura- 
cies on the order of 0.5mm would be required (about 1 
part in 68,000). On an RMS basis with 1/50th of a pixel 
pointing precision, this level was not quite achieved and