The impulse response and the modulation transfer func-
tion of the frame grabber, especially of the low-pass-fil-
ters which might not be adapted to the signal frequencies,
have several effects. On one hand the location of features
will be displaced due to the asymmetric impulse re-
sponse. Furthermore the internal precision will be de-
graded as the contrast of (small) targets is degraded due
to the reduction of the MTF. Figure 9 a shows an image
of a vertical line acquired with a MAX-SCAN frame
grabber without its selectable LPF. Figure 9 b shows the
same image acquired with the LPF. The asymmetric im-
pulse response and the ringing are evident.
A factor with significant importance is the performance
of the synchronization. Figure 10 shows the position of a
line in imagery acquired with PLL line-synchronization
(Figure 10 a) and with pixelsynchronous sampling (Fig-
ure 10b). PLL line-synchronization can induce a geomet-
ric deformation (in first order a shear) reaching several
tenth of the pixel spacing as well as linejitter. The shear
can be eliminated when HSYNC and VSYNC signals are
used for synchronization instead of the CSYN of com-
posite video signals. Pixelsynchronous sampling reduces
synchronization errors to below 0.005 pixel. Its stability
is a factor of 20 better than for PLL line-synchronization
via CSYNC signals. The random positional changes of
linejitter can be eliminated by averaging whereas the
geometric deformation cannot be eliminated through av-
eraging. It can on the other hand be modelled by includ-
ing a shear as additional parameter. It must nevertheless
be expected that pixelsynchronous frame grabbing pro-
vides for a higher accuracy than PLL line-synchroniza-
tion, whether or not that can be translated into a higher
three-dimensional accuracy must be verified.
D 3 T = 3 E
a) b)
- 0.69 [Pixel] - 0.02 [Pixel]
Figure 10 Position of a vertical line with PLL line-syn-
chronization (a) and pixelsynchronous sam-
pling (b).
2.7 Signal Transmission and Synchronization
Signal Transmission and synchronization are after tem-
perature the factors leading to the largest radiometric and
geometric disturbances. Image data transmission with
digital signals and analog signals using PLL line-syn-
chronization as well as pixelsynchronous frame grabbing
were analyzed. It could be shown that the geometric per-
formance of pixelsynchronous frame grabbing is identi-
cal to digital transmission and thus without loss. PLL
line-synchronization leads in connection with the deriva-
tion of the frame grabber synchronization signals from a
CSYNC signal to a shear reaching several tenths of pix-
els as well as linejitter. These effects were both deter-
mined via an investigation of the frame grabber and an
investigation using MEGAPLUS and SONY-XC77CE
cameras. It should be noted that linejitter does lead both
to geometric and radiometric instabilities. Methods for
the evaluation of the synchronization performance can be
found in Beyer, 1988 and 1991.
2.8 Temperature
The temperature plays a significant role for the internal
precision of measurements and such the accuracy to be
attained. Warm-up-effects occurring after a camera, and
to a lesser degree frame grabber, were switched on can
lead to large variations in image positions. These were
found to reach several pixels for PLL line-synchroniza-
tion. For a SONY-XC77CE a displacement of 0.005 pix-
el in x and 0.1 pixel in y was detected within the first
hour with pixelsynchronous frame grabbing. An internal
stability of better than 0.01 pixel in x and y after a maxi-
mum of three hours could be verified for periods of sev-
eral hours. The deformations occurring during warm-up
with pixelsynchronous sampling are attributable to ther-
mal deformations of the assembly. The larger variations
observed with PLL line-synchronization are connected to
variations of the clock frequency induced by the temper-
ature change during warm-up.
2.9 Averaging
The internal precision of positioning can be improved by
averaging. It was found that for pixelsynchronous sam-
pling an improvement corresponding to theoretical ex-
pectations can be achieved for at least up to 5 frames. It
could not be verified for longer sequences as the averag-
ing was performed off-line and only 5 frames were ac-
quired consecutively. If the time between series is several
minutes other (electrical and mechanical) effects limit
the improvement. It would thus be an advantage to use
real-time averaging over a number of frames for applica-
tions where a stable configuration is given.
3 THREE-DIMENSIONAL ACCURACY
3.1 Testfield and Network
The three-dimensional accuracy of a SONY-XC77CE
camera was investigated and verified using a three-di-
mensional testfield. The part of the testfield used in the
test contains 162 targets placed on a wall and on 5 alu-
