el-to-image co- 
yrientation ele- 
first and second 
yn were used as 
he third camera 
terior could not 
once by camer- 
ificant improve- 
cations are now 
e space is better 
erforming a spa- 
reference points 
is attained in X, 
accuracy in im- 
cing (0.52 and 
MS misclosures 
  
  
  
  
  
  
T | 15 
[um] | [ium] 
| 103| 191 
TL 18 
| 941 106 
| os7| 093 
13] 19 
  
  
  
  
| 3 are 0.69 and 
licates that a sig- 
ainable when an 
gorithms can be 
] precision mea- 
ersions is attrib- 
itations, some of 
timal as it had to 
raints. The con- 
rior and interior 
s an insufficient 
round the doors. 
icle led to a poor 
frames. 
all object points 
his was aimed at 
a high degree of 
ttained in the pi- 
ons of targets re- 
on and imaging 
IS imaged in 2 to 
the reliability of 
rably larger dis- 
ecurred at them. 
on the outside of 
targets inside the 
| a few positions 
Numerous targets were poorly imaged due to a number 
of reasons. The lighting was not properly adjusted, lead- 
ing to slight blooming. This is evidenced by the slight 
horizontal extension to the right for the targets shown in 
Figure 4 b. Figure 4 shows two views of the door and the 
Surface not 
flat 
a) Digitised 24 x 36 mm slide. 
b) Part of inverted frame of camera 1 
Figure 5 Images indicating some imaging problems. 
interior of the car. The image in Figure 4a was digitized 
from a slide print taken with a 24 x 26 mm film camera. 
The on in Figure 4b was grabbed from camera 1. Both 
were inverted for visualization purposes. Although the 
two images are taken from slightly different positions 
they show similar areas. Comparing the two images one 
can see for example that the targets in the door area are 
actually partially occluded (see the left circled targets). 
This cannot be detected due to the small number of pix- 
  
  
els on which they are imaged. Furthermore some targets 
are not on flat surfaces, which cannot be detected as well. 
Complete flatness of the targets will hardly be attainable 
in such an application as deformations of the targets dur- 
ing impact cannot be circumvented. These problems 
could be alleviated to some degree with “high-resolu- 
tion” cameras, i.e. cameras delivering over 1024 x 1024 
pixel, as occlusions might be detectable. 
The known difficulties indicate that the accuracy and re- 
liability can be improved. The measurement speed with 
the semi-automated procedure is already quite fast. The 
complete pilot test was conducted within 3 days. This in- 
cluded the set up of the test arrangement and the mea- 
surement of reference coordinates. This clearly shows 
that the integrated software package DEDIP is very pow- 
erful and flexible. 
5 CONCLUSIONS 
It was shown that the accuracy attainable with low-cost 
CCTV-type CCD-cameras is sufficient for many industri- 
al and other applications. The limited “resolution” of the 
cameras requires relative large targets on one hand, but a 
much larger number of images can be acquired and pro- 
cessed. This provides for a large enhancement of the reli- 
ability. Precise target location methods and positioning 
algorithms with suitable calibration techniques are a pre- 
requisite for meeting the accuracy requirements. Fully 
integrated software allows to meet the time critical pro- 
cessing demands. 
The test has thus clearly shown that Real-Time Photo- 
grammetry is able to meet a number of requirements for 
industrial and other measurement tasks. Important con- 
siderations to be addressed in future are the automation 
of the measurement process including the verification of 
the stability of calibration values and/or the exterior ori- 
entation elements. 
6. ACKNOWLEDGMENTS 
The author would like to thank all members of the de- 
partment “Konzernforschung - Vermessungstechnik” of 
Volkswagen AG for their fine cooperation and support in 
the pilot test. Acknowledgments are also extended to 
staff of the Institute of Geodesy and Photogrammetry for 
the realization of special camera and lighting fixtures re- 
quired for the project.