Calibration
Measurement
of system
Image coord.
calculation
3D coord.
Data
analysis
Results
| p2 | Dynamic param.
Figure 3. First order decomposition of the videometry system.
3.1 Measurement.
The aim of the measurement process is to obtain image
coordinates of the targets on the object for further use in the
object reconstruction. The choice of measurement method and
parameters determine the accuracy of the system and must be
regarded specially.
Image param
Image
grabbing
Signal
detection and
measurement
Image coordinates
Figure 4. Decomposition of the measurement process.
3.2 Targeting.
In most close-range applications artificial targeting is required
to produce sufficient quality and number of data points. The
need for artificial targeting was obvious for the teatcup liner
application since the surface of the liners are smooth and
uniform and do not provide any photographic contrast.
Two types of target points were considered for the teatcup
liners; both retroflective targets and projection of targets onto
the surface. When choosing the targeting it has to be considered
how the targets are imaged in the camera and the possibilities
of having homologous data points in the following calculations.
Other important criteria for the decision were speed of
execution and simplicity of procedure and hardware.
The use of retroflective targets is common in close-range use.
Circular targets of appropriate size are fixed to the surface in
a suitable pattern. This method ensures symmetric targets and
a best possible conformal representation of all the targets in the
image. This also forms the basis of successful target centre
determination.
Different methods of targeting with structured light were
considered. Structured light can be used both in an active
oriented sensor (active triangulation) and for target generation
(passive triangulation). A concept of regarding a raster projector
as an additional camera in a rasterstereographic system is
described by Frobin and Hierholzer (1981) among others. The
method has some advantages over conventional close range
methods, according to Ethrog (1991): a) Synchronization of two
cameras is not needed; b) Calibration of the projector with the
raster diapositive need only be done once; and c) When using
a linear pattern in the projection, the method is well suited for
automated evaluation.
Another variant of triangulation is Moiré technique based on
multistripe pattern projection (Tiziani, 1989).
Projection of structured light (i.e., laser) in either a stochastic
pattern or a regular raster, is another method of targeting. In
this method the surface is imaged simultaneously in two or
more images by usual photogrammetric procedure. For obvious
reasons, scanning arrangements are unsuitable for moving
surfaces.
The target projecting methods have the advantage of being
totally contact-free and not interfering with the measured
object. These types of targeting do not ensure homologue points
in subsequent images since there is no direct reference to the
object.
One precondition for success with such targeting is a bright
model surface. The rubber surface of the teatcup liner is dull
dark and made such targeting unfit for our case.
The choice for our prototype system was a regular pattern of
circular retroflective targets sized 9 1.5 mm. with intervals of
6.5 mm. This target size corresponds to approximately 5 pixels
in the image, which seems to be a favorable size for passive
targets (Trinder 1989).
Although this method implies targets fixed onto the surface, we
expect no influence from the targets on the characteristics of
the teatcup liner. We also obtain directly comparable results,
e.g. tension in the liner, from the measured target intervals.
Figure 5. A slightly flattened teatcup liner imaged in the left
and right image. The ellipses indicate located targets.
ath ^ aA
dA V. st
SAM
MN CN E es.