ell
for
are
ing
tic
yet
ion
ns.
ind
ICY
dle
ige
The measurement of image target location for orientation
was done with least squares template matching. The
camera orientation information was determined by a
bundle block adjustment, using 10 additional parameters
for compensating systematic image errors. All 12 frames
were used. Minimum datum information was applicd. The
adjustment gave the estimated image coordinate errors as
p, 7 0.50 pum (0.037 pixel) and p, — 0.61 jum (0.055 pixel)
with a system redundancy of r = 430.
The camera orientation parameters were introduced in the
edge matching as deterministic. The edge matching was
done with 4, 6 and 8 frames respectively. For the edge
tracking function a 3 pixel wide template ramp and a
patch size of 5 x 5 pixels were used. 173 edge points were
measured.
A straight line was fitted through the object points
projected onto the X-Y-plane and the X-Z-plane. The
differences between the observed coordinates and the
mean line (DXY, DXZ) are plotted in Figure 7 and their
minimum and maximum values and root mean square
errors (RMS) derived (Table 1). Additionally, the
standard deviations of the determined object coordinates
are plotted and their average values derived as
t2
t2
S. NS SY z- Y Sz
n n
and also listed in Table 1.
As evidenced by the RMS values of Table 1 the
deviations of the estimated knife edge points from an
adjusted straight line stay with 6 um well within the
1:25000 relative accuracy measure as far as the
horizontal distance (Y) is concerned. The respective Z-
(depth-) values exceed this specification, but they are not
of interest in the original application example. The
estimated standard deviations of the related object
coordinates, with values of about 14.5 jum in Y and 28.0
um in Z, are less precise, because they also include errors
in the orientation of the CCD-frames and are dependent
on the datum choice. Since in many inspection tasks the
accuracy of location and orientation of measured object
features is only of minor interest, these absolute precision
estimates are mostly of little value here.
This accuracy test and other tests with different set-ups,
camera configurations, templates etc. allow formulation
of some more general observations:
e If big differences between the template and the patches
occur, e.g. at corners, junctions and occlusions, the
| T T T T | T T T T | T T 1 | T T T T | T T T T |
0.040 [- 2
a A A tj
£)
0.024 = ^ -
E re a | v -
= Jason AO A +
£, 5 XO ETAGE SEES Sl apa OT A
N HT sn Re Rm RET CUT SN
€ 0.008[- my TEL ERU CER Hp ed or]
- HE CTE 2 oH
> 5 -
wn | | 1
> \ l t \
N —0.008[- \ l | : a ai}
a - prd pda Ley -
% he P ded j!
2€ - ini 4 ‘y I =
! Cp Ï
-0.024 - gu =
_ ly { J
{I -
—— — DxY - ' :
-----Dpx- u
LI — SX — Lu Ed | 1 1 5. 4 | 1 1 1 | 1 1 1 1 | 1 1 1 1 |
o =. SY i= 306 332 358 384 410
A — SZ - X [mm]
Figure 7 Edge matching results using the knife edge and six image frames. Differences DXY, DXZ
between estimated and straight line adjusted edge point positions. Standard deviations SX, SY,
SZ of estimated point locations.
