us
his
1S a
for
he
] in
; in
lay
| in
1 an
ity
ble
me-
In
und
and
and
and
epipolar line
template patch_1 patch_2
‚./images/tmpl ./images/s1_022
patch_3 patch_4
../images/s1_202 ./images/s1. 247
final position
initial position
Figure 4 Visualisation of the edge matching procedure using four images
the patch in the images. In the bottom row are the template
and the geometrically and radiometrically transformed
patches.
The matching algorithm is only usable for fine
measurement, because the LSM requires good
approximate values for the iteration process of the least
squares adjustment. The initial values consist of image
coordinate pairs of corresponding image points in all
images. The initial values must be delivered, e.g. by an
operator or automatically from a given CAD model.
Currently only the manual mode is implemented. The
images are displayed and the approximate image
coordinates of a corresponding point are measured with
the cursor. For measuring of more edge points the edge
tracking is used. The average pull-in range is half a patch
size.
In a possible automatic mode the initial edge points in
image space can be computed by resections of the model
edge point using the known camera orientation. In order to
exclude those images in which the edge points are
invisible a hidden-surface algorithm must be applicd.
4. ACCURACY TEST
The primary aim of this test was to verify the accuracy of
the edge matching algorithm. For this it must be well
understood that the accuracy of edge matching depends -
among other parameters (orientation, etc.) - on the
definition of the edge in object space, the contrast of the
edge and the edge strength in image space and the
amount of noise in the image data. Therefore it is of
utmost importance to have an almost ideal object edge for
accuracy testing available and to control illumination and
imaging in an optimum way.
The test object is a short knife edge (125 mm) with an
exactly polished edge for controlling the planarity of
surfaces. The technical specification for the planarity
tolerance of the edges is 2.5 jum (probability P=95%),
defined by DIN 874 (German Industrial Standard). All
points determined in object space should thus lie on a
straight line defined by the knife edge.
4.1 Set-up
To test of the algorithm a single camera/multiple frames
set-up was used (sce Figure 5). A CCD camera was
mounted on a optical bench construction and pointed
towards the object. The object was supported by a rotating
table, and fixed in a 3-D calibration field which was used
to determine the camera orientation. By rotating the table
an arbitrary number of CCD frames could be produced.
The calibration field consists of a black coloured plate (56
cm x 56 cm) and towers (10 and 30 cm height) with retro-
reflective targets. The targets are illuminated with a
specially designed fibre-optic lighting system around the
camera lens. The images are taken with a SONY XC77-
CE CCD camera with a Schneider XENOPLAN 1:1.7/17
lens and a Datacube MAX-SCAN framegrabber. The
framegrabber is controlled by a Sun-3E workstation.
From there the data is transferred via Ethernet to a
network of Sun workstations for further processing.
The framegrabber uses the composite video signal for
digitisation. The effective image size is 592 (H) x 574 (V)
pixels with a pixel spacing of 13.5 um (H) and 11.0 um
(V). At each camera position 4 images are acquired and
their average is used for further processing.
For the test 12 images are acquired: 8 images with a 45*-
interval table rotation and 4 images with a 90° rotation of
the camera about its axis (K rotation). Figure 6 shows one
of the images used for orientation and edge matching. The
hcight-base-ratio is h/b = 8.81 and the average image scale
is 1:44, i.c. 10 um in object space correspond to 0.23 jum
in image space or 0.017 pixcl with 13.5 um pixel spacing.
