Xt for example. Whilst 
currently only semi- 
nputing facilities are 
| power for complete 
ing the experiment if 
  
  
ipse. 
  
3 EXPERIMENTAL RESULTS 
Currently two geotechnical tests have been conducted 
using the described system to monitor soil surface 
change. In both cases these have been successfully 
carried out in conjunction with a conventional single 
camera system measuring displacements in the plane of 
the window. A typical image from the 2D system 
monitoring the front face of a model is shown in figure 2. 
Figure 3 is a typical image from the middle camera, of the 
set of three, viewing the top of the soil model. 
3.1 Computed target locations and precisions 
Since a bundle adjustment procedure has been 
employed, all computations include an in-built statistical 
analysis of the quality of all estimated parameters. For the 
each of the 3D tests thus far carried out about 940 targets 
were located in the soil surface. Of these about 90% were 
imaged by all three cameras. Typical rms standard 
deviations for the co-ordinated targets are 0.057mm, 
0.042mm and 0.098mm in X, Y and Z axes respectively. 
The lower precision in the Z direction is expected since 
the camera configuration was limited by mechanical 
constraints within the centrifuge such that all optical axes 
were parallel to the Z axis to within about 20 degrees. 
Whilst the quality of these results is not as good as that 
obtained with displacement transducers, the large number 
of target points provides a set of dense information 
spaced evenly over the soil surface. Unfortunately 
displacement transducers could not be used to provide an 
independent check as their use would have obscured the 
Soil surface. 
3.2 Surfaces of change 
Figure 4 shows the soil surface computed from the set of 
images corresponding to figures 2 and 3 where the model 
has all but collapsed. It should be noted that the depth 
axis (Z) has been exaggerated in this figure. Figure 5 has 
been computed by simple differencing of the grid 
generated from images at the beginning of the experiment 
and that from figure 4. These data, and others from the 
sequence, are currently being analysed in conjunction 
with 2D movements computed from images taken with the 
single camera viewing the front of the sample. Of 
particular interest are any discrepancies between the 
movements in the soil surface close to the window and 
those measured at the window by the conventional 2D 
system. 
4 CONCLUSIONS 
8). A 3D-measurement system able to monitor change in 
surfaces of geotechnical models tested on a centrifuge 
has been successfully designed, constructed, calibrated 
and tested. 
b). Initial results appear promising and will undoubtedly 
enhance understanding of geotechnical events. 
C). With the availability of low cost computing and imaging 
hardware, there is considerable scope to further develop 
and automate the technique. 
d). It is expected that on-going developments in both 
automation and visualisation of information will allow both 
2D and 3D monitoring to take place on-line when real-time 
feedback is required by the engineer. 
787 
5 REFERENCES 
Allersma, H.G.B., 1991. Using image processing in 
centrifuge research. Proc. Centrifuge ‘91, Balkema, 
Rotterdam. pp 551 - 558. 
Cooper. M.A.R., and Robson. S. 1996. Chapter 2: 
Theory of close range photogrammetry. In Close 
Range Photogrammetry and Machine Vision. Edited 
by K.B.Atkinson. Whittles, Caithness, Scotland. 371 
pages. ISBN 1 870325 46 X. 
Ethrog. U. 1994. Strain measurement by video 
image processing. Proceedings of the 10th Recent 
Advances in Experimental Mechanics, Silva Gomes 
et al (eds) p411-415. Balkema, Rotterdam. ISBN 90 
5410 3957. 
Fraser, C. S., & Shortis, M.R. & Ganci, G. 1995. 
Multi-sensor system self calibration. SPIE Vol. 2598: 
Videometrics IV, pp 2-33. Philadelphia, USA. ISBN 
0-8194-1962-1. 
Preparata, .F.' P. & Shamos, M..]|, 1985. 
Computational Geometry: An Introduction. pp.203- 
248. Springer-Verlag, New York. ISBN 0-387- 
96131-3. 
Robson. S., Clarke. T.A.,, & Chen J. 1993. 
Suitability of the Pulnix TM6CN CCD camera for 
photogrammetric measurement. SPIE Vol. 2067. 
Videometrics Il, Boston. pp. 66-77. 
Robson. S., and Shortis, M. HR. 1997. Surface 
Characterisation from tracking discrete targets. 
Optical 3-D Measurement Techniques. Edited by 
Gruen and Kahmen, pp 179-186 Pub: Wichmann, 
Zurich ISBN 3-87907-311-2. 
Shortis, M.R, Clarke, T.A. & Robson, S. 1995. 
Practical testing of the precision and accuracy of 
target image centring algorithms. SPIE Vol. 2598: 
Videometrics IV, pp 65 - 76. Philadelphia, USA. 
ISBN 0-8194-1962-1. 
Taylor, R.N, Robson, S., Grant, R.J. & Kuwano, J. 
1998. An image analysis system for determining 
plane and 3-D displacements in soil models. In 
press for: Proc. Centrifuge '98, Tokyo, September 
1998. Balkema, Rotterdam.