Standard deviation (mm)
Post spacing (pixels)
Figure 5 - Standard deviations of the differences between
R-Wel DMS and Zeiss P3 DEM’s, for different
post spacings and correlation matrices.
As already mentioned, it was decided to use convergent
photography for this work because it was expected to give
better results. Photography with no convergence was, however,
also taken with a base to distance ratio of 1:2.7 and initial tests
show that the standard deviation increased as expected. For
example, for a 5 post spacing DEM with 17x17 correlation
matrix, the value is 1.42 mm and this is also 2.3 times the pixel
size.
These initial results have been achieved without a full
calibration of the system. No account has been taken of lens
distortion effects or of any other random geometric effects in
the CCD chip. These will be investigated in further work.
Several other aspects also need further consideration before a
working system is complete:
e it will be essential to employ a synchronised pair of
digital cameras when dealing with a living subject.
This is seen to present no real problems, except
possibly cost.
e it will be necessary to ensure that the shutter speed
used is as fast as possible to arrest any movement
whilst at the same time any projected pattern must
remain visible. The current exposure time of around
0.25 secs. is possibly too long for a patient to remain
still.
5. CONCLUSIONS
The requirement for medical practitioners to demonstrate the
effectiveness of any treatment is increasing and with this will
come a greater demand for measurement to support particular
methods of treatment. This work has investigated the potential
of a low cost digital photogrammetric system, which offers a
degree of automation, to provide an - effective medical
measurement tool. Such systems are capable of being used by
medical practitioners themselves, which was not the case with
analytical and analogue systems, and this could be a decisive
factor in extending the use of photogrammetric measurement in
medicine. Early indications suggest that such systems are able
to provide many of the measurements needed to the required
accuracy and within the desired time schedule, and further
work is in hand to confirm these findings.
REFERENCES
Fanibunda, K.B., 1983. Photoradiography of facial structures.
British Journal of Oral Surgery, 21, pp. 246-258.
Gabel, H., Wester-Ebbinghaus, W., Woytowicz, D., Hallbauer,
Th. and Schumpe, G., 1992. Photogrammetric measurement of
the human back shape and its relation to the spine.
International Archives of Photogrammetry and Remote
Sensing, 29(B5), pp. 862-866.
Graham, R.W., 1995. Kodak digital cameras for small format
aerial photography. Photogrammetric Record, 15(86), pp. 325-
327.
Mitchell, H.L., 1994. A comprehensive system for automated
body surface measurement. ^ International Archives of
Photogrammetry and Remote Sensing, 30(5), pp. 265-272.
Thomas, P.R., Mills, JP. & Newton, I, 1995. An
investigation into the use of Kodak Photo CD for digital
photogrammetry. Photogrammetric Record, 15(86), pp. 301-
315.
Welch, R., 1989. Desktop mapping with Personal Computers.
Photogrammetric Engineering and Remote Sensing, 55(11), pp.
1651-1662.
410
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
KEY WOF
ABSTRA(
We have c
The CMU
of 30 millic
240 pixels
bility of prc
for 3D visi
Stereo rar
of two or |
great dea
stereo sys
throughpu
mapping.
effectively
depth me
range of
the densit
ter the dy
are sever:
1990, Wel
yet none
output of
latency.
We have
has the tl
put transl
speed of
depth res
depth m
machine
baseline
and Kana
multiple ir
different E
Video-rate
is passive
With appr
lution ca
range im
range inf«
common
scanless;
shape dis
sor suffer
video-rate
applicatio
Z keying,
time.
