An example of running locus measurement is shown
in Figure 5. The specified turning radius (15m) and
measuring results are shown in this Figure. The
measured locus of 5 times running are overlaid in the
figure. The measured values represent the locus of right
front edge of the vehicle.
As for the measuring result of running locus, we
could obtain the measured values within 0.5m from the
specified radius in general for any turning radius. The
measuring results were generally satisfactory excepting
that discontinuity took place in the locus for about 10 cm
when the system switched over the image for
correlation.
(2) Comparison of the results taken at different
observation altitude
In this study, the same pattern of test run was
observed from two different altitude of 100m and 150m.
We could obtain almost similar results both from 100m
and 150m altitude photographing because this extent of
difference could be offset by the adjustment of zoom
lens of video camera. So ,it may be better to select the
high observation altitude so as to minimize the
displacement of image.
(3) Influence of mid-way stop at turning
The car was stopped (for 2 seconds) at the mid-way
of leftward turn in this experiment. We confirmed
whether the deviation took place during the stop, and
* Measurement reference line
Turning radius 15m
Figure 5 Example of running locus measurement result
(Observation altitude 100m, Turning radius 15m )
382
found that no significant deviation occurred to the locus.
We could also obtain the graph to show the change of
running speed including those of mid-way stop. As a
result, we could confirm that mid-way stop does not give
any significant influence on the measurement in case of
our measuring system.
From the above, we confirmed that appropriate
results are obtained in general in the measurement of
running locus. Partly, there were cases where matching
was impossible because the road surface and vehicle
could not be distinguished each other just like in the
cases of speed measurement, therefore, we are
planning to add pre-processing to photographing
condition to make it robust.
6. Application for the Identification of Car Type
We attempted to identify the car type utilizing the
measuring results of running speed. If it is possible to
classify the running cars into the rough two groups of
normal size car and large sized car, it will be quite
effective in saving energy in the current traffic
investigations.
In this experiment, we calculated length of cars
passing through the test section using the following
formula, and discussed the possibility to identify the
type of car by the difference of length.
T
L=2> Nenn WI.
n=1
Where, L: Car length;
W: Length of measuring section
v: Measured Speed of car (each 1/60 second)
t: Unit time of measurement (1/60 second)
T: Time when the car passed the section
n: Number of video field
When this method was applied to the video image
used for the measurement of running speed, the
measuring error of maximum 2.5m was observed in
case of normal size white car and maximum 1.5 in case
of large sized car. In case of the car with black body, the
measuring error close to 4.5m was observed at
maximum. It is barely possible to distinguish normal
size white passenger car from large sized car as there is
a difference of 3m or more in the length between the
normal size and large sized cars, but the judgment is
quite unstable. The main cause of these errors would be
the accumulation of error involved in measuring the
speed at the unit of 1/60 and measuring error of passing
time of the section. The reason why the error is larger in
case of black car body is attributable to the fact that car
body cannot be distinguished from road surface of
asphalt at the automatic tracing by image matching.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
From
employ :
contours
Asa
the mea:
and phot
on road s
by the im
simple re
value p
distinguis
patterns
car was
necessar
of road s
method,
correspo
video im
extract tl
the extra
of vehicle
the expel
tracing €
Figur
