ABSTRACT
As the social economic activities grow variously in recent days, an advanced traffic control system corresponding
to the changes of road traffic circumstances becomes an urgent matter. And traffic information service or the profit
convenience for travelers and commodity distribution are also needed.
April in 1993 under such situation, the synthetic and epoch-making traffic management system, Universal Traffic
Management Society of Japan(UTMS), was established. Among the purposes of UTMS, especially Integrated Traffic
Management System(ITCS) aims at the development of a more optimum signal control algorithm which can corresponds
to the every moment changing traffic situation. So to realize this, an advanced sensor which can collect more precise and
more detail traffic flow information is needed.
September in 1993, UTMS established a new working group, the purpose of which was research and development
of an image sensor which provides useful traffic flow information for the traffic signal control which realizes this ITCS.
And in this working group we started examining a specification of the image sensor.
This image sensor processes images received from the ITV camera installed above the approach lane at the traffic
signal intersection, and can directly measure queue and delay length up to 150m as well as number of passed vehicles,
speed, vehicle type and vehicle existence in 30m area.
So we have developed new measurement algorithms and a new hardware, and we succeeded in the development
of a new image sensor, SPatial Image processing Traffic flow Sensor(SPITS), which satisfies this specification.
While the function of direct measurement of queue and delay length is the advantage of an image sensor, it is hard
to measure them precisely in the distance from the camera position because of poor visibility. SPITS has succeeded in
enabling to measure queue and delay length precisely even in the distance with a sophisticated method and in real time
processing.
SPITS has been evaluated by field tests. The experimental results show that SPITS has achieved accurate
measurement within 10% error of maximum queue length under various conditions(day, night, twilight, rain etc.) and
confirm that SPITS is able to provide useful traffic flow information for ITCS.
In this paper we will show these results as well as the queue and delay length measurement method used and
hardware configuration.
CHARACTERISTICS OF AN IMAGE PROCESSING SENSOR
As the sensor which collects traffic flow information, mainly Ultrasonic Vehicle Detector has been used. This
detector detects vehicle presence by the time difference of the reflection of ultrasonic wave fired from above the road
surface to just under it. But especially queue and delay length are measured indirectly by the number of passed vehicles
in a unit time. So a sensor which can collect more precise traffic flow information is needed. Also each Ultrasonic
Vehicle Detector has to be installed above the road surface per a measurement lane and so there is a fear of spoiling the
beauty of the city.
On the other hand, the Image Sensor processes an image received from the ITV camera installed aside and above
the approach lane at the traffic signal intersection, so it is able to directly measure a spatial area like shown in Figure 1.
And the Image Sensor has the following characteristics.
(1)It is able to measure speed, vehicle type as well as number
of passed vehicles on more than one lane at the same time.
(2)It is expected to collect more precise and more detail
queue and delay length information because of the direct
measurement of them.
(3)It does not spoil the beauty of the city so much because
one image sensor installed aside and above the approach lane
can measure more than one lane at the same time.
With these characteristics, the image sensor is thought to be
one of the main sensors at present.
DEVELOPMENT GOAL
In October 1993, we started to develop an image processing
traffic flow sensor taking an advantage of image sensor's
characteristics and matching to the purpose of UTMS. So we started
to develop an image processing sensor which can measure number of
passed vehicles, speed, vehicle type, vehicle existence, queue length, and delay length directly on more than one lane at
the same time.
Especially, in case of the queue and delay length measurement in the distance from the camera position, the
vehicle on the image is so small that we need an advanced measurement technique to measure them correctly. And we
had to measure all these measurement items at the same time and in real time. Also it was necessary to deal with the
coming to the practical use.

Download full report
http://utms.or.jp/english/inter/paper/paper02.pdf