position where CAD diagram and stereo image accord the most 
with each other. What’s more, though each point in CAD 
diagram has a coordinate precision of less than a decimal point 
after perspective projection has been made, it is cut down to an 
integral coordinate value on the frame memory ( 512 pixels X 
512 pixels ). Thus, a gray level cannot be obtained at the coor- 
dinate precision of less then 1 pixel. Therefore, it has been 
decided to obtain the gray level at the coordinate precision of 
less than 1 pixel from the gray levels of 8 pixels nearby the 
pixel of each point in CAD diagram by interpolation. The 
worker can judge a degree of conformity between CAD dia- 
gram and stereo image with the sum of gray levels displayed by 
the monitor as an index. Fig.8 shows a monitor image obtained 
by superimposing CAD diagram on the stereo contour image of 
flange. This system incorporates such feature that since a three- 
dimensional translation amount is given directly to CAD data, 
it is unnecessary to determine a corresponding point which 
becomes necessary in a stereo photogrammetry. 
  
Fig.8 Stereo Contour Image 
4.3 Measurement accuracy 
With a flange of 350mm in diameter taken as an object of 
measurement, the center position and posture of the flange was 
measured by the measurement system of total station using the 
foregoing target and the measurement accuracy was confirmed 
by CAD base measurement system with the above measurement 
results used as CAD data. The flanges taken as the measure- 
ment object was photographed as large as possible in the visual 
field of camera at two video total station. In this case, 1 pixel of 
the frame memory is equivalent to about 1mm in actual size. 
Table 3 shows the translation amount given till the sum of gray 
levels at each point in CAD diagram becomes maximum. 
  
  
  
  
Table 3. 
X (mm) Y (mm) Z (mm) 
Position 0.4 0.2 0.3 
Posture 0.1 0.2 0.1 
  
  
  
  
  
  
The value shown in Table 3 becomes a deviation between the 
measurement system of total station using the target and CAD 
base measurement system. The deviation between both systems 
was maximum 0.4mm in a translation amount and maximum 
0.2° in a rotation amount. It has been verified from the results 
that the measurement accuracy secured by CAD base measure- 
ment system satisfies a desired accuracy even if the accuracy 
insured by the system of total station using the target is taken 
into additional account. 
5. Conclusion 
Aiming at promoting the pipe prefabrication method in a 
chemical, a three-dimensional measurement system has been 
built up with the flange of apparatus and pipes in the plant 
taken as the measurement object. With application of a system 
using a total station requiring a target at the measuring point of 
the site, it has been confirmed that the prefabrication method 
can be realized by the three-dimensional data control. Further, 
in order to save the time and labor required for mounting the 
target, the three-dimensional measurement system was exam- 
ined by collating a stereo image with CAD diagram which is 
the measurement object and a system including 2 servo-drive 
type video total station has been built up. In particular, it has 
been confirmed that a problem of correspondence detection 
encountered in the stereo measurement can be worked out by 
applying the concept of CAD base vision which is advocated 
now in the field of image recognition and also, a desired meas- 
urement accuracy can be satisfied by collating the sub-pixel in 
CAD diagram with a contour image obtained by image process- 
ing, and a prospect of putting the system to practical use has 
been obtained. 
Acknowledgement 
Thanks are due to Dr. H. Chikatsu of Tokyo Denki University 
for helpful suggestions in CCD camera calibration and to Mr. 
M. Tanaka of SOKKIA Co., Ltd., for valuable information 
about total station. 
Reference 
Fishler,R.B., 1987. Model Invocation for Three Dimensional 
Scene Understanding, Proc. 10" IJCAI, pp.805-807 
Shunji,M., 1981. A Study on Analytical Photogrammetry with 
Use of Non-metric Camera, Report of the Institute of Industrial 
Science the University of Tokyo, vol.29, No.6, July 
356 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
  
  
KO] 
ABSTR 
In indu 
delivere 
project 
is a nee 
like poi 
photogr 
projecti 
KURZI 
Industri 
als CA 
Mebsys 
macht 
notwent 
Orienti 
welches 
des Ver 
Im indi 
metrie 1 
Verfahr 
Zu tret 
zeichne 
untersc] 
photogr 
punkte 
unmitte 
digitale 
Dieses 
on i.d. 
Koordii 
hende i 
Input f 
durch s 
beschre 
Modell 
werden 
der Foi 
Design. 
Herstel]