the locus. 
change of 
stop. As a 
es not give 
| in case of 
ippropriate 
Jrement of 
> matching 
nd vehicle 
like in the 
, we are 
tographing 
‚ar Type 
ıtilizing the 
possible to 
groups of 
Il be quite 
ent traffic 
th of cars 
e following 
dentify the 
) second) 
cond) 
tion 
ideo image 
speed, the 
)bserved in 
1.5 in case 
ck body, the 
bserved at 
lish normal 
r as there is 
yetween the 
judgment is 
rs would be 
asuring the 
r of passing 
r is larger in 
fact that car 
surface of 
'atching. 
  
From the above résults, it is considered better to 
employ another method such as the extraction of 
contours of vehicle to identify the type of car. 
7. Issues in the Future 
As a result of experiment, it has become clear that 
the measuring accuracy expected from CCD camera 
and photographing altitude can be realized if the pattern 
on road surface and vehicle can be clearly distinguished 
by the image analysis. On the other hand, in case of the 
simple recognition method of car body such as binary 
value processing, it was almost impossible to 
distinguish a car body from the marks and complicated 
patterns of road after rainfall, and automatic tracing of 
car was difficult in many cases. Therefore, it is 
necessary to realize robust processing for the condition 
of road surface. As an improvement for the processing 
method, we are now studying the method to implement 
corresponding points retrieval to the total area of two 
video images which are temporally continuous, and 
extract the contours of moving object. Figure 6 shows 
the extraction result by improved method of the contours 
of vehicle for which automatic tracing was impossible at 
the experiment this time. It is considered that automatic 
tracing of car body will become far more stable by 
  
(b) Extracted contours of cars 
Figure6 Extraction result of the contours of car 
383 
applying this method. The problem is that the hardware 
which can process this method on real time has not 
been developed so far. Since the algorithm has already 
been known, it is desirous to incorporate into hardware. 
If the contour of vehicle can be extracted, it will be 
possible to distinguish normal size car from large sized 
car by measuring the area inside the contour or 
calculating the circumference of contour. It is expected 
to lead to the improvement identification rate comparing 
with the measurement of car length which was 
attempted in this experiment. 
8. Conclusion 
In this study, we made basic discussion on the 
method to automatically measure the image taken by 
video camera mounted on balloon so that monitoring of 
traffic flow and measurement of vehicle movement can 
be easily implement at optional place. 
Using the video camera mounted on balloon, the 
vehicles traveling below and the GCPs placed around 
the road are recorded as video image. By the video 
image analysis made on computer, matching based on 
image correlation and automatic tracing are made 
simultaneously to vehicles and GCPs to be monitored. 
The image coordinates obtained as a result of automatic 
tracing are converted into actual length, from which 
running locus and running speed of vehicle are 
calculated. 
We made experiments to verify the measurement 
accuracy of this method. As a result, it was possible to 
make satisfactory measurement if the running vehicles 
and road surface could be distinguished on the image. 
On the other hand, there are some cases where 
matching happened to be defective and automatic 
tracing was impossible because of the marks painted on 
the road or disorder of color of road surface resulting 
from rainfall ,etc. To cope with this problem, we are 
currently making improvements to provide robust 
measurement by changing pre-processing of matching. 
References 
Yamana,R.,Yahara,T.,Mori,M.,Setojima,M.(1995):Auto 
matic Measurement of Traffic Flow by Video Camera 
and Unmanned Balloon. In: Proceeding of Annual 
Conference of Japan Society for Photogrammetry and 
Remote Sensing. pp. 221-224. 
Iwasaki, Y., Sadakata, A. (1989) Measurement of 
Space and Traffic Flow using a Kite Balloon, Kyushu 
Tokai University Information Center ,Japan. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996