the accuracy
>S compared to
ler accuracy is
tical axes which
camera) than in
oerpendicular to
tion of camera
e and geometry
lies, the sensor
of 0.015 pixels.
th six LEDs and
depth (z) of 200
sults for different
o Z [mm]
n]
0.05
0.18
0.41
0.79
1.25
2.65
ults for SCS
perimental mea-
shows the results
amera to object
nm] | e Z [mm]
5 0.05
6 0.13
5 0.26
38 0.54
58 0.87
results for SCS
‘simulated and the
ion of one of the
perimental results
This indicates that
3} on the sensor
ditions.
st was done to find
its. 50 points were
| better than 0.001
optical axis of the
ely parallel with the
were fitted to a
mathematical plane and the measurement standard
deviation was estimated based on the residuals. For a
.camera-to-object distance of 2500 mm, the accuracy was
found to be (1*o level);
Distance [mm] | o, [mm]
2500 0.010
Table 3: Plane measurement results for SCS
It is clearly shown that the accuracy in the directions
normal to the line of sight (X and Y direction in the simula-
tions and experiments) is superior to the accuracy parallel
with the line of sight (Z direction). With other words, the
SCS is a metrology system where the angular
measurement accuracy is superior to the distance mea-
surement accuracy.
More extensive tests, like standardized CMM accuracy
tests, have not yet been done. Nevertheless, both the
simulations and the experimental results so far indicate
that the SCS meets the accuracy requirements for a
variety of industrial metrology applications.
5. Post Processing of SCS Measurements
If on-line measurements are not of paramount importance,
accuracy can be significantly improved by repeating mea-
surements of the same object points using different
camera positions, and subsequently entering the
observations into a post processing module. The accuracy
improvement achieved by the post processing approach
is based on utilizing the high angular accuracy of the SCS
System. The post processing approach is ideal for the
establishment of high precision reference networks.
As explained in Section 3, the SCS software is based on
estimating the position of the Light Pen tip as referenced
to the camera coordinate system. When the Pen tip is
pointing to a measurement point, the point of contact is
mathematically back-projected on to the sensor, based on
the estimated relationship between the coordinate system
of the Light Pen and the camera coordinate system (see
Figure 1). Equation 1 is used for this perspective
transformation. The back-projected point constitutes the
"fictitious" observation that is recorded and subsequently
input to the post processing bundle adjustment. For every
camera position, the same object points (triangulation
points) are touched with the Light Pen tip, and the
"fictitious" sensor observation is calculated for each of the
object points. The accuracy of the back projected point is
determined by the angular measurement accuracy of the
SCS.
In the post processing bundle adjustment, the 3D
coordinates of the triangulation points are estimated. The
accuracy of the triangulation points is dependent on the
geometry of the photogrammetric network. Network
geometry is characterized by the number and orientation
of camera stations, and number and position of
triangulation points. Due to the approach of "indirect"
measurement of triangulation points using the Light Pen,
the points are observable from "all" directions. Therefore
it is easy to achieve a favorable network geometry when
operating the SCS.
The full 3D measurement capability of SCS provides
approximate coordinate values for all the unknown
triangulation points, which is needed for the post
processing. To achieve a high accuracy, the post
processing is a free network bundle adjustment only
constraining given distances between some of the
triangulation points.
6. Applications
To obtain a favorable result for SCS applications, the
accuracy characteristics of the SCS must be taken into
account. Two main application groups can be defined:
1. Applications where full 3D measurement
capabilities are not required, like
straightness measurements.
2. Applications with low accuracy requirements,
where the moderate length measurement
accuracy of SCS is satisfactory.
Some examples include:
= Measurement of the straightness of airplane
fuselages. If the line of sight is approximately
parallel with the airplane fuselage when
doing the measurements, the high angular
accuracy will provide a high accuracy of the
straightness measurements.
= Measurement of wing contour. If the line of
sight is parallel to the wing surface, high
accuracy is achieved.
= Measurement of flush and gap on nacelle
(aircraft engine cover). Flush is measured
with the camera aimed in the length direction
of the nacelle, while gap is measured with
the camera pointing perpendicular to the
length direction.
= Measurements on collision tested cars. This
is an application where the accuracy
requirements usually are moderate.
Measurements inside the crash tested car
are particularly easy to perform.
= Establishment of high precision reference
networks by SCS in combination with post
processing.
References
[1] Alf Pettersen. "Metrology Norway System -
Optimum Accuracy based on CCD
Cameras". International Archives of
Photogrammetry and Remote Sensing,
Vol.XXIX, ISPRS 1992, Commission V.