perspective transformation. In this process, the 
apparent movement of cars in the image due to the 
vibration of balloon is eliminated, and real moving 
distance will be identified. 
3. Method of Experimental Measurement 
We implemented experiments to take video image 
and measure the test vehicle in order to verify the 
accuracy of the measurement made by computer. 
As for the test run section, we specified each one 
section separately for running speed measurement and 
running locus measurement. 
For the measurement experiment of running speed, 
we specified straight section of the length about 20 m on 
the normal traffic road, and shot the video image of the 
test vehicle running through the section. At the same 
time, we recorded the actual running speed by 
measuring the passing time of the test vehicle using a 
stop watch. We used 2 normal size passenger cars and 
one large sized car as the test vehicles. Of these normal 
size passenger car, one has a car body of white color, 
and the other had the black color. Also, we took video 
images by changing the altitude of balloon to 3 different 
levels of 50m, 100m and 150m. The running speed of 
test vehicle was specified at 4 speeds of 15, 30, 45 and 
60 km/h, and the test was made for 5 times for each 
speed. The number of test run is 70. 
In the measurement experiment of running locus, we 
turned a car to the left at the radius of 5m, 10m and 15m 
on the test section where the left turn had been assumed, 
and took the video image of it from the upper air by the 
balloon mounted video camera. The altitude of balloon 
at the time of photographing was 100m and 150m. Also, 
we considered the movement of vehicle at the 
intersection. Namely, we specified two patterns; one is 
to stop for two seconds in the mid-way of left turning and 
then re-start to continue the turn, and the other is to turn 
the corner to the left without stopping at the same speed. 
60 cases were measured in total. We used normal size 
passenger car of white body color as the test vehicle. 
4. Measuring Results of Running Speed and 
Discussion 
As for the accuracy of measuring result of running 
speed, we compared the values of actual measurement 
and value of automatic measurement from the following 
viewpoints. 1)Comparison by the kind of cars, 2) 
Comparison by running speed ,3)Comparison by 
observation altitude 
(1) Comparison of measuring results by the kind of 
cars 
In this study, the image correlation was calculated by 
converting the video image into binary value in order to 
implement matching processing on real time. The: 
measuring accuracy was compared by the type of car 
because it was expected that the classification of car 
body in binary value would vary according to the type of 
vehicle. Figure 3 shows the difference between actual 
speed and measured speed based on the video image 
  
  
  
  
  
  
8 
: $ 
E 
E 4 E--MMMRMÓeeeeG $a $e 
S 2pee,yá o6 Sere 
E 8 
zm : 
t $ 9 
= Sn ese vuiniviein s . a e a a Aie Que 
2 cd fe-serepetemne cent 1201000080 eM $ Mee 
e 
= a A SE an A eR ere A EE 
-8 E 
White body Black body 
Normal size car 
Large size car 
Figure 3 Measuring error of running speed by kinds of cars 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
  
taken at € 
The ce 
and large 
other kin: 
matching 
body colo 
from the ¢ 
and matc 
(2) Com 
spee 
Figure 
an examj 
The m 
groups. C 
the actue 
having tl 
groups hi 
matching 
having la 
test run, 
on the rc 
because 
road surf 
In cas 
differenc 
the actu 
different. 
vehicle. 
camera 
measure 
km/h of t 
those wif 
ft 
Là 
Measured value(Km/h)