a fully automatic target coordinating system would 
be completed in a matter of minutes. In terms of 
dealing with clients this is almost 'real-time', and it 
would certainly be acceptable to have a client wait 
for internal consistency checks on the acquired 
targets, and to take another image if required. 
The target coordinates would be acquired much 
more cheaply. In addition to the cost of film, 
processing, and printing, the traditional approach 
would take some 2 hours to observe with sufficient 
care. 
The CCD approach has the potential to be a much 
more predictable and reliable provider of target 
coordinates. To achieve this one would need to 
obtain comprehensive calibration data on the cam- 
era/lens combination, and to have a particularly 
consistent targetting system. Under these conditions 
one could be very confidant of achieving predictable 
precisions of target coordinates - not necessarily the 
same over the whole format. Precision variations 
are usually less of a problem for adjustment than 
unpredictable (i.e. unknown) changes in precision. 
As mentioned earlier in this paper, the cost of electronic 
componentry tends to reduce at a significant and steady 
rate. This is true for CCD cameras, and especially true for 
frame grabbers, image processing software, and correspon- 
ding computer hardware and software. These savings are 
partially offset by increasing power and facilities being 
offered at all levels (CCD array size, frame grabber 
capacity, sensitvity, and speed, image processing capability, 
and computer speed and memory capacity). So we can 
identify another advantage for CCD cameras; 
4 
Reducing cost, from an intially high figure. It is 
inevitable that 512x512 pixel systems will eventually 
be cheaper than 70mm film cameras. Because the 
latter will always require highly skilled manufacture 
and assembly their cost will remain relatively static 
whereas all levels of CCD array systems can be 
expected to become cheaper in real terms. 
We should note that where the CCD camera is used to 
scan conventional photography the advantages are limited 
to numbers 2 and 3. Because a conventional film camera 
and photographic processing are involved in the observing 
procedure the turnaround time and the cost of the system 
will be worse (or hardly better) than a straight photo- 
graphic system. 
5 
  
Depending on the array size required, a CCD array 
camera would be much less bulky than a 70mm film 
camera. If this is pleasing to the shoemakers and 
their clients it is a consideration. However the 
system requires the inclusion of a computer to 
  
   
capture the digital (digitised?) image, which rather 
tips the 'bulk' issue in the opposite direction unless 
the computer is located and controlled remotely. 
DE-MERITS OF CCD ARRAY SYSTEMS 
For the time being CCD array systems have demon- 
strably lower image resolution than (say) most 
70mm film cameras. While this has clear implica- 
tions for stereoscopic observations, it will not always 
matter directly in the acquisition of target 
coordinates. Shortiss (1991) has reported target 
coordinate accuracies of 0.03 to 0.07 pixels for small 
CCD array cameras, which is well within the re- 
quirements of this project. The limitations could 
come with target images of low contrast, so system 
design, especially illumination and contrast of 
targets, will require special attention. 
However Trinder's work (Trinder, 1989) indicate 
that in order to achieve the sort of positional 
accuracies quoted above the target diameters should 
not be less than 5 pixels. Beneath that level the 
accuracy of the pointing drops off very quickly. 
Now if one were to require targets to have an image 
diameter of 5 pixels and a minimum spacing 
between targets of 10 pixels (with a 512x512 pixel 
array). There are 11 places on my initial trial where 
the targets are closer than this rather optimistic 
minimum spacing, and this was with a regularly 
shaped foot. Where the foot being measured had 
significant deformities there could be some dificult- 
ies in defining the shape with sufficient resolution. 
In order to achieve the minimum target size noted 
above, then sperical targets would need to be 10mm 
in diameter. These would be dramatically large 
targets to place on the client's foot. They are likely 
to be as unhappy as I am at the prospect, and this 
is no way to introduce people to a new process (that 
they will probably be paying for!). 
For the time being it seems that the cost of camera, 
frame grabber, and software is still notably higher 
than for a 70mm film camera (for example) and 
good tablet digitiser. Even now this extra cost 
needs to be balanced against the time saved. If a 
working system were to measure 100 feet in a year, 
and save 2 hours digitising on each one, then the 
annual savings over a photographic approach would 
be 200 hours of skilled time, plus the photographic 
processing costs. Over a 4 year 'life the CCD 
system would save 800 skilled hours, plus processi- 
ng, plus a 70mm camera tablet digitiser, less some 
reasonable share of the development costs of 
software to acquire the target coordinates. This 
equation can be evaluated for any particular place.