49
8.2.3 Fiducial Marks
The image co-ordinate system is defined by fiducial marks in the camera
frame, which was filed to form small corners. There are eight marks in each
camera, and their positions define the image co-ordinate system. The co-ordi
nates of the fiducial marks were determined in the following way. A series of
exposures towards a bright and unfocussed background was made in each
camera. The fiducial marks in these photos were measured in a comparator
and then transformed to one and the same system. Because of film shrinkage,
setting, rounding-off etc the transformed co-ordinates differ between the pic
tures. (see formulas 8.3.1 and 8.3.2)
On each picture all measurements were repeated to estimate the standard
deviation sj, which contains the effects of setting errors, rounding-off errors
and accidental errors inherent in the comparator.
The variation of the co-ordinates of the fiducial marks between the pictures
gives us a possibility of estimating a standard deviation, s 2 , which also contains
the effect of film shrinkage and possible errors introduced by the camera during
exposure.
The co-ordinates of the fiducial marks are the averages of all transformed
measurements and the precision of the averages is calculated from the standard
deviations of the transformed co-or-dinates. Testing the equality Si = s 2 , the dif
ference between Si and s 2 is highly significant. The standard deviation of the
averages is 0.58 and 0.59 jum for the two cameras. The number of degrees of
freedom is 480 and 528 respectively. The standard deviation of a single deter
mination of the fiducial co-ordinates is 3.17 jum and 3.42 jum respectively.
8.2.4 Fiducial Transf ormations
The calibration of the stereocamera is based on three pairs of pictures of the
test object, no. 1 and no. 2 with aperature f:45, and no. 3 with f: 11. The first
Table 7. Fiducial transformations. Stereo-Medical-Nikkor.
Pair of pictures
Cam 1
Cam 2
Scale
S scale
Scale
S scale
1
.999622
.001214
1.001074
.000290
2
.999818
.001214
1.001129
.000220
3
.999684
.001244
1.000728
.000354
Mean
Variance between
Variance within
.999708
10036 10-12
.001217
1481087 10-12
1.000977
47257 10-12
.000293
85849 10-12