DISCUSSION ON THE AUTOMATIC MEASUREMENT OF TRAFFIC FLOW 
BY VIDEO CAMERA MOUNTED ON BALLOON 
Masaru MORI, Masahiro SETOJIMA, Yukio AKAMATSU 
Environment division ,Kokusai Kogyo Co., Ltd. Japan 
Ryo YAMANA, Takashi YAHARA 
Chugoku Technical Institute, Ministry of Construction ,Japan 
IWG V /TI 
KEYWORDS: Vision , Correlation, Matching , Real time, Video ‚Traffic, Speed, Running Locus 
ABSTRACT: 
It is necessary to objectively monitor the running speed and running locus of the vehicles traveling on 
existing road to solve the traffic congestion or to design the intersections. Monitoring by video camera is 
effective for this purpose, but it has a drawback in installation cost and selection of installation places. In this 
study, we attempted to use unmanned balloon photographing as the traffic measurement system so that 
monitoring can be done easily at any optional place along the roads. Namely, a video camera was mounted 
on balloon and the balloon was sent up to take the video image of traffic flows just below the balloon. We 
applied image matching continuously to the video images to identify the running speed and running locus of 
individual cars in the image. The magnitude of tremble of video image due to the vibration of balloon was 
measured by applying image matching together with the vehicles. Then, we obtained the real running distance 
of vehicles by subtracting influence of tremble from the movement of vehicles. In order to evaluate the 
measuring accuracy, we acquired the video image and measured the test vehicle which traveled at pre- 
determined speed and locus. As a result, it has become clear that excellent results are obtained both for 
running speed and running locus only if the automatic tracing of vehicle by stereo matching is successful. In 
some cases, however, significant measurement error was observed when automatic tracing was difficult 
because of the influence of color or pattern of the road. With this respect, we proposed a method to improve 
  
the accuracy of automatic tracing of vehicles in this study. 
1. Introduction 
It is necessary to monitor and analyze the current 
state of road traffic in order to solve the problems such 
as ever increasing traffic congestion and accidents. 
Installation of industrial television (ITV) is now under 
way on main national roads and highways in Japan, and 
the monitoring system by video image is partly in 
operation. 
However, excepting the places where permanent 
monitoring is necessary, installation of video camera 
has been difficult because of the high cost of 
construction. Also, it is difficult to find out a location at 
suburbs to install camera to shoot the road from proper 
altitude and at proper angle because there is few tall 
buildings along the roads. 
It is necessary to measure the running speed, 
running locus and kinds of vehicles to implement 
various kinds of traffic plan and management such as 
design and improvement of road shape, traffic control to 
mitigate congestion and road repairing plans. At the 
places where the video images taken from high altitude 
are not available, only the possible way of measurement 
is the qualitative observation made by human eyes at 
the road side. Therefore, it is desirous to develop a 
measuring system which can shoot video images of 
road from high altitude at any places. Also, it is desirous 
to develop a computer based automatic measurement 
system to measure the speed and locus of moving cars, 
because the number of vehicles to be monitored is so 
large. 
In this study, we mounted a video camera on 
unmanned balloon, and took the picture of road . Then 
we attempted to identify the running speed, running 
locus and kinds of each vehicle in the video image by 
applying image processing thereto. We applied image 
matching for measuring the location of moving cars and 
for automatic tracing of vehicles. As for the apparent 
movement of cars resulting from the vibration of balloon 
caused by the winds at upper air, we worked out the 
movement of camera by automatically tracing the 
ground control points(GCPs), and made corrections to 
the movement of cars. 
This paper reports about the results of accuracy 
verification test of the methods described above and the 
development of improving methods. 
378 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
  
  
(1) Obser 
As sh 
on an airs 
5.2m and 
21m? of 
equipmer 
balloon h: 
the heigh 
The vi 
integratec 
inch. Ler 
direction 
operation 
composir 
direction 
In phc 
about 1r 
more pos 
these cor 
video frar 
(2) Meas 
As the 
measure! 
measurin 
Individua 
equipmet 
Unr 
B: 
  
Fic