raveling on 
) camera is 
aces. In this 
tem so that 
as mounted 
balloon. We 
ing locus of 
balloon was 
ng distance 
valuate the 
2led at pre- 
ed both for 
ccessful. In 
vas difficult 
| to improve 
| eyes at 
levelop a 
nages of 
| desirous 
surement 
ving cars, 
red is so 
mera on 
ad . Then 
, running 
image by 
ed image 
| cars and 
apparent 
of balloon 
d out the 
icing the 
actions to 
accuracy 
/e and the 
  
2. Configuration of observation/Measurement 
System 
(1) Observation system 
As shown in Figure 1, we mounted a video camera 
on an airship type balloon. The total length of balloon is 
5.2m and maximum diameter is 2.6m. When filled with 
24m? of helium gas, it can lift up the photographing 
equipment of around 5 kg to the altitude of 250m. The 
balloon has two mooring ropes, which are used to adjust 
the height of balloon according to the wind condition. 
The video camera mounted on the balloon is deck 
integrated type VHS-C camera. The size of CCD is 1/3 
inch. Lens is 7 - 41mm zoom lens. Photographing 
direction and zooming can be controlled by wireless 
operation from the ground. The number of pixels 
composing the video image is 660 for horizontal 
direction and 400 for vertical direction. 
In photographing the road, we placed circle mark of 
about 1m in diameter as ground control points at 6 or 
more positions and adjusted the camera so that both of 
these control points and vehicles can be confined in a 
video frame. 
(2) Measurement system 
As the basic study for the automatic system of traffic 
measurement based on computer, we employed the 
measuring method as shown in Figure 2 in our study. 
Individual function were realized by image processing 
equipment and EWS. 
  
Unmanned 
Balloon 
  
  
  
  
Video Camera 
  
  
  
Figure 1 Diagram of the observation using 
unmanned balloon 
      
After unloading the video tape on which the pictures 
were taken at upper air, the tape is replayed and stored 
as digital image in the frame memory device. At present, 
it is possible to record the data for only 8 seconds, but 
the moving images for longer time can be continuously 
stored on the computer memory if digital video, etc. will 
come to be in use more commonly in the future. 
6 or more ground control points are specified in the 
first frame of video image which is composed of plural 
number of frames. Also, the initial position of the 
vehicles to be monitored is specified in the image. In the 
future, we are considering to automatically distinguish 
the vehicle bodies from background(road) and to 
mechanically select the vehicles subject to the 
automatic tracing. At present, the number of vehicles 
which can be specified and traced simultaneously in one 
image is around 10. 
As for the ground control points and cars for which 
the initial position was specified, the image coordinates 
are measured sequentially in the continuous frames at 
the frequency of 1/60 second. In this study, we used 
Quick Vector made by OKK Inc. (Japan) for real time 
image matching. 
Image coordinates of 10 cars and 6 ground control 
points (GCPs) are outputted to and recorded in 
engineering workstation every 1/60 second. Eventually, 
actual length conversion factor is calculated from map 
coordinates and image coordinates of ground control 
points, and position coordinates of cars are outputted in 
the form conforming to the map coordinates using 
Video image 
Frame memory 
( Specification of GCPs’ initial position ) 
  
  
  
   
  
  
  
  
( Specification of initial positions of vehicles | 
  
  
( Tracing by stereo matching ) 
  
  
(output of position coordinates | 
  
  
(Real length conversion, position correction | 
  
  
(Running speed, running locus] 
  
Figure2 Flow chart of measurement 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996