deflection 
axis, were 
ally plotted 
32. 
ersection 
Z 
0.70 
  
  
  
  
ITION 
ric system 
capable of 
le to sub- 
J region of 
root of the 
ay painted 
face glare 
o-reflective 
mounted 
dinated by 
9 provide a 
‘gets were 
sure of the 
  
  
resolution 
itor 
tem 
re used to 
strates the 
> structural 
| calibrated 
e arrays at 
ar, far ang 
> then used 
  
Camera 4 
8.5mm lens 
  
measure a 
    
Figure 5 An image from camera 1(8.5mm lens). 
to provide principal point offset, radial and tangential lens 
distortion parameters. 
The turbine blade was statically loaded in 10 increments. 
At each load three sets of five convergent images were 
captured. Unfortunately specular reflections from the fibre 
glass blade surface under the bright laboratory conditions 
meant that target location routines developed for optimum 
conditions could not locate all necessary targets. An 
additional set of images without axial illumination were 
made at each load increment. Subtraction of this image 
from each retro-target image considerably reduced the 
background allowing identification of the target image 
points. However this method has the effect of increasing 
image noise, so the target locations were simply used as 
seed points for target measurement in the conventional 
retro-target images. In this way the imaged targets were 
automatically measured and identified on site to sub-pixel 
accuracy using a centre weighted algorithm (Clarke at al, 
1995). Target image coordinates from the three sets of 
data acquired during each epoch were averaged before 
processing. 
4. RUDIMENTARY ON-SITE VISUALISATION 
Given the limited data processing power of the current 
image acquisition and processing system, it was not 
11 
Figure 7 Image movement vectors between epochs 10 
and 11 at a scale of 10x. 
  
  
495 
    
Figure 6 An image from camera 2 (16mm lens). 
possible to reliably compute object space co-ordinates 
on-line during the experiment. Recent research work in 
the field of geotechnical image measurement and 
analysis (Chen et al 1996) has resulted in some rapid on- 
line target measurement and movement vector overlay 
procedures. By levelling the central camera it is possible 
to compute and draw image space movement vectors 
which are useful to the engineer. These can be overlaid 
onto the central image for on-line viewing. Figure 7 shows 
the image movements occurring between epochs 10 and 
11. From the figure it can be seen that targets on the rotor 
blade have moved with respect to the camera, whilst 
control targets separate from the blade exhibit no 
movement. A display of this type can be extremely 
reassuring during a costly one-off experiment, it can also 
provide valuable information to the structural engineer. It 
is also possible to retrospectively analyse movements 
between any two image measurement sets, for any 
camera station. This can be very useful in checking the 
performance of each camera and can provide warning 
that a camera has been moved or knocked during the 
experiment so that new starting values can be computed. 
Figure 8 shows similarly derived image movement vectors 
computed between epoch 10 and 12 where the load was 
increased to 175% and the angle of the jig supporting the 
rotor blade changed to provide sufficient ground 
     
  
Add 
  
Figure Image movement vectors between epochs 1 
and 12 at a scale of 10x. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996