IF slip rings 
  
  
  
  
  
  
  
clay samplé with 
targets on 
viewable surfaces 
^". 
     
   
   
   
. camera 
"(front view) 
model 
container 
swirlging 
platform 
  
perépex 
window 
Figure 1. The arrangement of the geotechnical centrifuge, detailing the camera configuration used. 
as the geotechnical experiment progresses. The 
conventional method is to use linear displacement 
transducers to measure change at discrete locations on 
the soil surface. Whilst these can provide very precise 
measurement (better than 0.025mm) of the soil surface at 
each discrete point, it is not feasible to measure sufficient 
points to define the soil surface. This paper describes an 
alternative method, which uses an image based 
measurement solution to measure the 3D movements of 
several hundred targets located at the soil surface. 
2. ELEMENTS OF THE VISION METROLOGY SYSTEM 
The 3D measurement procedures used to determine initial 
soil surface shape and its subsequent change are 
founded upon the capture of image sequences from three 
synchronised CCD cameras to provide overlapping views 
of the targeted surface of the soil sample. Common target 
images in the three sets of image measurements obtained 
are identified and measured before use within a 
photogrammetric procedure to compute the 3D co- 
ordinates of the targets at the soil surface. These co- 
ordinate data can then be triangulated to determine the 
shape of the soil surface for each set of images. 
2.1 Image sequence acquisition 
A gantry structure has been designed to support three low 
cost 2/3” monochrome CCD cameras at accelerations of 
up to 100 g. The cameras employed in this case are a 
single Pulnix TM6CN and a pair of Pulnix TM6EX 
cameras (Robson et al, 1993). Each is fitted with a fixed 
focus 'C' mount lens of 4.8mm focal length. The analogue 
image data from the three cameras monitoring the soil 
surface are passed out of the centrifuge environment to a 
784 
host computer by means of slip rings. Whilst the Pulnix 
cameras have not been designed to operate at 100 
gravities, the applied forces can be simply regarded as a 
self-weight problem. As such the camera mountings are 
designed to securely support the heavier lens and "C" 
mount assembly with the cameras being operated within 
20 degrees of verticality. To date no problems arising from 
the high gravity environment have been experienced, 
although they were expected and procedures were 
adopted to detect them. 
To monitor effectively dynamic change in the soil surface 
the cameras must be synchronised. To achieve this the 
first camera provides a set of master image control 
signals that are then used to drive the second and third 
cameras. In this way the set of three monochrome images 
can be sent to a colour frame grabber as if they were the 
red, green and blue image planes from an RGB colour 
camera. In-house image capture software then treats the 
resultant colour image as three colour bit planes and splits 
these back into three monochrome TIFF images for 
subsequent storage to disk. Based on a Matrox Meteor 
PCI frame grabber, running under Microsoft Windows NT, 
the system is able to store images to hard disk at a rate of 
up to one set of three frames every two seconds. This 
method allows sufficient image storage for all typical 
geotechnical tests which may run for anything up to 24 
hours. In this system hard disk transfer provides the 
bottleneck, however the frame rate achieved is sufficient 
for the purposes of most geotechnical events. 
2.2 Datum definition 
To ensure correct scale and orientation of the computed 
soil surfaces it is necessary to establish a datum. To this 
  
  
Fi 
end