4.2 Bullet Train (Shinkansen)
Then we proceeded to actually measure the nose of
Shinkansen train by automatic measurement. The
measurement was carried out in two different
methods. First, it was made by a stereo-model
from a single position. The other was made by a
stereo-model moving through 5 different positions
with control points to make successive adjustment.
The Fig 11 shows 5 different positions of
stereo-model and the checking points. To assess
the accuracy of the checking points and to measure
the control points we used IMS(Industrial
Measuring System) MS-200. This system is to
measure the 3D coordinates with triangulation,
using the theodolite and subtense bar. Its relative
accuracy is 1/8000.
The table 1. shows the results of each of the two
different methods.
As to the requested and targeted accuracy of
1mm (Cf:(1) equation), the table 1.D shows that
we have attained it fully.
In the case of table 1. 2 , as we did not change
the camera's position on the base-bar and moved
back the apparatus from the object, the distance
became greater and the measuring area was
narrowed, thus impeding the measuring plan to
meet the requested accuracy. So, the accuracy was
lowered, although the targeted accuracy was still
satisfied.
Although we did not try this time, if we make a
plan of measuring the area of 3000mm X 1000mm
within the targeted accuracy Imm, we are
confident that it is possible to attain such
precision even with the measurement by 1
stereo-model from a single position.
MODEL-1 MODEL-2
MODEL-3
MODEL -4
MODEL-5
oe alla ns
1600
Fig.11 Positions of image-acquiring part and checking points
The time spent between camera operation and the
end of analysis was about 1 hour for 1
stereo-model measuring from a single position
(6000 points spaced 25mm from each other);
whereas for the measuring from 5 successive
positions it took about 7 hours (28000 points
spaced 15mm from each other). In near future with
the progress of algorithm and speeding-up of
computer, we believe that it will become possible
for 1 stereo-model operation to take only 15
minutes for 6000 points.
S. MEASUREMENT DATA
MANAGEMENT AND APPLICABILITY
TO CAD SYSTEM
PI-1000 can output the 3D data of analyzed result
in the format of DXF file. DXF file is a standard
data format, which can be input into almost any
CAD systems and makes possible the mutual data
usage between PI-1000 and other CAD systems. So,
we believe it can also be applied to contrast
composition of CAD data and design data.
The Fig.12 is a perspective view of 3D data of
table 1. D. This perspective view was produced
successively 1) first by changing the 28000 3D
data obtained from PI-1000 into wire-frame, 2)
second by inputting the data into AutoCAD
through the DXF format, 3) third by applying to
them rendering process on AutoCAD and 4)
fourth by giving them 3D view angles. The Fig.12
shows very clearly the exhaust port and light
section of the train and even the slight wrinkles of
the paper covering the light are visible in an
enlarged picture.
$T Li
15713, 7. 559 Bon AEF QA EN
Fig.12 Perspective view (AutoCAD)
Table 1. Measuring conditions and measured results
Again
produce
table 1
CAD.
The
process
process
output
image -f
be inpu
the ma
example
this tim
The P
by 3D
stereo .
device:
data, as
of surf
the spo
The Fi
of the
process
Further
data on
image
compos
Thoug
with Sh
it at the
In this
outline
simulat
real tra
measure
test anc
results
we coul
such d
differen
image w
Our sy
comple»
Number of | Measuring | Distance Base checking | Control Target Resulted Point Total
models area (mm) length points points accuracy | éccuracy number time
(mm) (mm) (mm) (mm) in DTM
D5models 3400 X 1900 800 33 12 1. 0 0.6 28188 Thours
1600
2 1model 2100 X 3000 800 27 0 2.7 1.8 6561 lhour
2000
80
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
