ind
nal
ge
ed
2),
Jh.
1S
al
al
i
Fuse, Takashi
A New Technique for Vehicle Tracking on the Assumption of Stratospheric Platforms
Takashi FUSE" and Eihan SHIMIZU"
" Department of Civil Engineering, University of Tokyo
” Professor, Department of Civil Engineering, University of Tokyo
Hongo 7-3-1, Bunkyo-ku, Tokyo, 113-8656
E-mail: <fuse, shimizu>@planner.t.u-tokyo.ac.jp
JAPAN
KEY WORDS: Object Tracking, Movement Detection, High Resolution Data/Images, Image Sequence, Platforms,
Algorithms, Stochastic.
ABSTRACT
Traffic flow survey for traffic control and planning is usually conducted with traffic beacons set up at limited roadside
points. Therefore, they cannot observe the exact dynamic movement of vehicles which is important information for
sophisticated traffic policy. On the other hand, stratospheric platform system has been recently projected in Japan.
One of the purposes of the stratospheric platform system is utilization for earth observation. The stratospheric
platform is expected to result in high spatial and time resolution images at specific areas for continuous observations.
These high resolution and continuous images certainly make observation of vehicle movement easier. In this paper,
we explored the possibility of vehicle tracking with high resolution and time-serial aerial images, which are on the
assumption of the use of stratospheric platforms. In estimation of displacement vectors of vehicles, the most
characteristic problem is that appearance/disappearance of vehicles occur, when they are under overhead bridges or
shadows of buildings, or going outside the image, or so on. We employed the probabilistic relaxation method for
tracking vehicles. And then we improved the probabilistic relaxation method by introducing (1) the color information
of vehicles , and (2) the displacement vectors of each other. We applied the proposed method to simulated data and
sample images, which were on a one-way street. The time interval of successive images was 1.5 seconds. The
proposed method yielded a better result than the original method, and the rate of correct correspondence is above 95%.
Furthermore, we also applied to the various time interval images. When time interval was less than 1.5 seconds, the
result was good for vehicle tracking in this case.
1. INTRODUCTION
Traffic flow survey for traffic control and planning is usually conducted with traffic beacons set up at limited roadside
points. Therefore, they cannot observe the exact dynamic movement of each vehicle which is important information
for policies decision for traffic problems. There have been a few attempts made to study the survey of dynamic
movement of vehicles, for example, with aerial photographs. The results, however, were not sufficient for useful
application to traffic engineering.
In recent years, a stratospheric platform system has been projected mainly by the Science and Technology Agency and
the Ministry of Posts and Telecommunications in Japan. It is basically intended to contribute to telecommunication
and the other different purposes. One of them is utilization for earth observation. The stratospheric platform is being
supposed to be kept at a stratospheric altitude of about 20km, so it is expected to result in high spatial and time
resolution images at specific areas for continuous observations. These high resolution and continuous images certainly
make observation of vehicle movement easier. As a result, the stratospheric platform has a great potential with wider
scope of utilization, for instance, survey for taking countermeasure against traffic jam, for taking origin-destination data,
survey of rate of right and left turn in intersections for signal control, and so on.
In this paper, we explore the possibility of vehicle tracking with high resolution and time-serial aerial images, which are
on the assumption of the use of stratospheric platforms. To be specific, we develop a new technique for vehicle
tracking with high spatial and time resolution remotely sensed images. We propose improved probabilistic relaxation
method as the new technique by introducing (1) the color information of vehicles, and (2) the displacement vectors of
each other. And then, we confirm the effectiveness of the proposed method through applications to simulated data,
sample images and different time interval successive images.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 277