A Method for counting pedestrians using a Laser Range Scanner
ABSTRACT
Video-based approaches have been studied for decades in order to count passing people.
However, these methods can be successfully applied only when the cameras can be located at
such a good position that they can collectively minimize occlusions and can cover an entire area
with sufficient resolution. In addition, image processing algorithm for counting people using a
video camera has not yet overcome the difficulties such as high computational costs, occlusions
by the dense pedestrians. Except when only smaller numbers of people pass through the gate at
the same time, the video-based methods are difficult to count the pedestrians very accurately in
a relatively larger space. In this paper, we propose a method of counting pedestrians using a
laser range scanner. A laser range scanner is mounted on the ceiling of a passage, and scans
vertically keeping the scanning plane orthogonal to the major passage direction, i.e. the laser
beam scans the passage like a curtain. We can obtain the height values of passing people with
the laser scanner with 37.5 Hz frequency. The height data from the laser scanner fluctuates as
the pedestrians passed there. Furthermore, even if the multiple pedestrians are passing
simultaneously, the laser range scanner can discriminate the human shapes individually because
of a high angle-resolution. Accordingly, even if the density of pedestrians is so high, we can
count the number of passing people accurately by counting the fluctuant laser range data or
height values. An experiment was conducted in a railway station in Japan, which was used by
250,000 passengers a day, and in the observation site, the passing pedestrians were about 240
passengers per minute in the rush hour. Through the experiment, the performance of the laser-scanner-
based counting system was demonstrated.
1. INTRODUCTION
Analyzing or monitoring human activities, such as counting the number of passengers, or
measuring their trajectories, is considered very useful in various fields such as building security,
planning and management support for shopping mall, railway station and so on. So far, motion
analysis and counting pedestrians with video data has been a major method. Video-based
systems for people counting over time and detecting overcrowded situations in underground
railway stations were described in (Regazzoni and Tesei, 1996). A lift aiding system by
counting the number of passengers waiting at each floor was developed in (Schofield et al.,
1997). In order to reduce occlusions, video cameras have to be carefully located due to limited
viewing angles and resolutions. For video-based surveillance, the change of illumination and
weather conditions can be one of the major obstacles to its reliabilities. On the other hand, laser
range scanner (briefly called “laser scanner” in the followings) using eye-safe laser have been
successfully applied through many research efforts to improve efficiency and reliability in
obtaining object geometry and detecting suspicious persons. In recent years, single-row type
laser scanner with high scanning rate, wide viewing angle and long-range distance has been
developed, and can be bought with rather low price on market.
In this research, we propose a method of counting pedestrians automatically using a laser
scanner. The experiment was conducted in a railway station in Japan, which was used by
250,000 passengers a day, and in the observation site, the passing pedestrians were about 240
passengers per minute in the rush hour. The results are compared with the number of pedestrians
counting by visual observation with the height data.
