376
Figure 3. Root end of wing with measurement adapters (photography courtesy of British Aerospace Airbus).
The performance of the systems had to be proven in numerous trials (a task repeated with every customer) before the systems were
allowed to be used in routine inspection tasks. The results of some of these tests shall be used to demonstrate the geometric accuracy
performance of the still-video cameras and the IMETRICS software.
A first test was performed using a “bolt” model - a master model of the wing root. The targets spanned a volume of 1.0 x 3.2 x 1.3m.
Stick-on targets, button targets and special measurement adapter targets, developed by the client in conjunction with IMETRIC, were
applied. This resulted in 119 targeted points. The employed retroreflective targets have a small central dot which is used for theodolite
observations. This set-up served, with other tests, to perform a basic comparison between the industrial theodolite system and the
IMETRICS system. Two sets of measurements were taken with both systems. A two theodolite configuration was utilised for the
measurement task with the ECDS system. A network of 20 images was generated using a Kodak DCS420 camera. The measurement
conditions were less than ideal as a large change in temperature, exceeding 5 degrees Celsius, occurred between the start and finish of
the measurements.
Both bundle adjustments of the IMETRICS TP210 system showed RMS image coordinate residuals in the order of 1/70th to 1/100th
of the pixel spacing. All the data sets were compared using a 3D spatial similarity transformation. The results of the comparisons are
given in table 2 below.
Version RMS X (mm) RMS Y (mm) RMS Z (mm)
ECDS(1) vs ECDS(2) 0.039 0.040 0.030
TP210(1) vs TP210(2) 0.041 0.038 0.017
TP210(1) vs ECDS(1) 0.071 0.084 0.049
TP210(2) vs ECDS(2) 0.067 0.086 0.049
Table 2. Results of comparisons of “bolt” model.
In the table:
RMS : Root mean square of the residuals.
ECDS() : Data set of ith measurement with the ECDS system.
TP210(1) : Data set of ith measurement with the IMETRICS TP210 system.
The results showed a good internal agreement between the two sets of each of the systems as well as between the two systems. The
relative accuracy of the comparisons corresponded to 1 part in 98 000 for the ECDS system, 1 part in 106 000 for the IMETRICS
system, and 1 part in 52 000 for the comparison between the two systems. The reduction in the agreement between the two systems
was in part attributed to the eccentricity of the targets used by the respective systems.
In another test the repeatability of the system was checked using a standard measurement task (see figure 3). For one wing type, the
root and volume check requires the coordinates of 156 targets to be determined in a 0.5 x 2.8 x 1.0m volume. To study the
repeatability of the photogrammetric system, image sets with 10 and 45 images were acquired and measured by different operators.
Three operators acquired 2 sets of 10 images each and one operator acquired 2 sets of 45 images each. The acquisition of the images
was spread over 3 hours. This was to assess whether environmental changes within such a time frame would degrade the accuracy or
not.
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1 995
(y). Th
^ 405. Ul ".
te
l