Camera position
acquisition
MEASUREMENT
istics of train body
For example, often
out design, or sandy
lustrous, or three
ve, many of which
cult to measure by
| preliminary testing
S to see whether our
ally works and then
iin body.
oints on a simulated
f 400mm X 300mm
0mm and with the
ained a satisfactory
mm(rms) and target
imm (Kochi, Ohtani,
| to test- measure the
d surface of 300mm
sion in the center to
ly the light from its
led on the different
the right and left
difference in 'the
ed images of two
periment it did not
z much to... our
>ction obtained from
processed through
iracy of our system
| the result obtained
rement-apparatus of
joints on the same
the base length was
of 1048mm. The
sfactory with the
i. (rms) in depth
ith the targeted
nna 1996
The targeted accuracy is calculated by the
following equation.
8 Z-(H?X 6 p)/(fx B) (1)
ó Z:target accuracy
(depth)
ó p:accuracy of
reading the image
coordinates
H:distance between
camera and object
B:base length
f :principal distance
Fig.6 Simulated surface
os oem
Fig.7 Cross section (PI-1000)
3.2 Application on a Stainless Train
We experimented our system on a stainless train
body at Niitu Train Factory of East Japan Railway
Co.Ltd. We measured the section 800mm X
800mm below the window. The measuring
conditions were the same as 3.1. The Fig.8 shows
the actual scene of measuring. Fig.9 shows the
perspective view obtained by the automatic
measuring processed by PI-1000.
As the stainless body is lustrous, it affects
strongly the shading on the right and left images,
thus forcing us to narrow the measuring area down
to 600mm X 600mm. Nevertheless, with the
exception of its periphery we could obtain a
satisfactory data of 3D. The requested accuracy
was 0.5mm, but we actually obtained the accuracy
of 0.1mm (rms), fully satisfying the requested and
targeted accuracy.
Fig.8 Scene of measuring
79
Fig.9 clearly shows a stripe-form protrusion of the
train surface as well as the targets purposely
placed thereon.
The mismatching on periphery was due to the
fact that while the illuminating light of pattern
design was almost saturated near the center (255
level), the amount of light decreased through
shading down to almost 0 level, as it went to the
periphery . In order to solve the problem, we are
now developing the CCD camera (presently 256
level) of greater Gray scale resolution, thus
widening the dynamic range.
Fig.9 Perspective view (PI-1000)
4. MEASUREMENT OF NOSE CURVE
Basing on our research results so far obtained, we
now proceeded to measure the 3 dimentional and
larger surface. We first tested, therefore, on the
simulated surface and then experimented on the
actual nose of a bullet train (1000mm X 3000mm
X 1000mm) of East Japan Railway Co.Ltd.
4.1 Simulated Surface
We made stereo-matching automatic measurement
experiment on two simulated surfaces. One was a
white flat plate of 1000mm X 1000mm. The other
was a similar plate now curved with radius rate of
600mm. We measured two of them together
simultaneously in the same measurement area,
which was, therefore, 1000mm X 2000mm with the
base length of 2000mm from the distance of
3000mm.
The Fig.10 shows the perspective view realized by
our PI-1000 from the measurement results. The
flatness and the curve are clearly observed.
Fig.10 Perspective view (PI-1000)
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
