| point
nt of each
maximum
53
86
34
88
8
63
29
for all sets
he principal
cipal point
al distance
igher than
much more
it for CLC3
As it is shown in Fig. 6 for the Fuji S1 Pro the correction grid
covers the fact of having non-squared pixels using a resolution
of 2304 by 1536 pixel. The correction grid is based on a grid
width of 2mm. The pattern of points represents the image
measurements with respect to the sensor. The plot shows the
reproducibility of the correction grid of one camera. The
marginally difference results from slight variation of image
point spread. The correction values concerning the DCS 460
indicate no significant deformations.
Figure 6: Correction grid for CLC1 and CLC2- 100x inflated
3.3 Exterior accuracy
For the evaluation of the photogrammetric adjustment the
exterior accuracy related to reference distances is finally to be
estimated. In this case 120 distances, derived from 16 reference
points, are computable. The reference coordinates are only
measured once and underlie high variation of temperature and
transport influences from the CMM to the photogrammetric
laboratory. To evaluate the influences by temperature, constant
measurements of temperature are performed during and before
image acquisition. The results of the photogrammetric system
are corrected with respect to the co-efficient of expansion. The
system scale is based on one scale in y-direction. The scale has
an a priori accuracy of Sum (calibrated by PTB).
For CLCI-3 Fig. 7 shows the resulting errors of length
measurement (LME). CLC1 and CLC2 are trend-corrected with
30um/m, CLC3 is trend-corrected with 40um/m. Obviously a
length-dependent part for the adjustment result is remaining.
Beneath this general trend a continuity concerning the length
measuring error (LME) with respect to the discrete reference
points is visible. AII LME for distances with point 209 remain
negative and form minimum in LME. Some other effects by
reference points are remaining equally for all data sets.
Having a view on the collectivity of LME for each data set an
absolute mean of LME (6) and a RMS value for LME (7) can be
determined.
7
LME, = à abs (6)
n
2 2 2
s? = > ty + e "s + X s +
Ox, ro 2 oy, 7
These values are taken as expected deviation for LME. 67 —
97% of all LME are captured within 2c, respectively +40um in
object space. The CLC3 taken with the Kodak DCS 460 shows
high spread in LME. This probably results from the careless
ZU
n
RMS, Me = (7)
Table 4 shows the maximal corrected deviation in positive and
negative direction and mean values of LME.
Max. deviation LME RMS; me
CLC +0.0629 -0.0530 0.0170 0.0217
CEC2 +0.0450 -0.0410 0.0129 0.0160
CLC3 +0.0683 -0.0960 0.0346 0.0423
Table 4: Analysis of length measuring error
For further analysis the lo, 20 and 3c standard deviation for
one distance computed of 2 object points (8) are calculated for
the additional observation that implies the system scale (Table
5, the %-value points the LME within the Sigma value).
length measuring error- CLC 1
0.140
0.120
0.100
0.080
0.060
0.040
0.020
0.000
-0.020
-0.040
-0.060
-0.080
-0.100
-0.120
-0.140 s : . Histance ....
trend-corrected 30um/m
deviation in mm
length measuring error- CLC 2
0.140
0.120
0.100
0.080
0.060
0.040
0.020
0.000
-0.020
-0.040
-0.060
-0.080
-0.100
-0.120
-0.140 distance
trend-corrected 30pmim
deviation in mm
length measuring error- CLC 3
0.140 -—
0.120
0.100
0.080
0.060 Soie +
= 00404? 24* d Wah
E 0.020 3 + > = + , e. 3 +800
= 0.000 > “+. * 4e. : %
s 0 r T$ $ +
© -0.020 288 #88 588 2 +005
: ch e* Se 33 ,
= -0 ++ 1
-0.060 +— + ste i
+ + +
-0.080 ———+* A
-0.100 2. $$
-0.120
-0440 + 4 distance
trend-corrected 40umim
Figure 7: Diagram of length measuring error
6s ts; + zs sts lass (8)
Oy; 2 Oz, 2 Oz, 7
handling and the fact of using an instable camera, in this case
the CCD-array is only fixed with one screw.
281-—
