Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 86
Edited by: B.H.V. Topping
Paper 75

Traffic Calculations using Stereo Video Cameras

H. Yoshida1, H. Muraki2, S. Tanaka3, H. Furuta3, Y. Nishita4 and S. Fujimaki5

1Graduate School of Information Technology, Kobe Institute of Computing, Hyogo, Japan
2ImageONE Co. Ltd., Tokyo, Japan
3Faculty of Informatics, 4Graduate School of Informatics, Kansai University, Osaka, Japan
5Asia Air Survey Co. Ltd., Kanagawa, Japan

Full Bibliographic Reference for this paper
H. Yoshida, H. Muraki, S. Tanaka, H. Furuta, Y. Nishita, S. Fujimaki, "Traffic Calculations using Stereo Video Cameras", in B.H.V. Topping, (Editor), "Proceedings of the Eleventh International Conference on Civil, Structural and Environmental Engineering Computing", Civil-Comp Press, Stirlingshire, UK, Paper 75, 2007. doi:10.4203/ccp.86.75
Keywords: traffic flow of people, image processing technique, congestion, 3D coordinates, cluster analysis, lap probability.

It is necessary for determining the maintenance plan of a road to predict the traffic in the future by analyzing the current situation of the regional traffic and the regional environment. In order to assess the current traffic volume, a road traffic census is temporally placed into effect. The result of the investigation is applied to a design and management plan for roads and it is used for as background material for setting various road policies.

Until now, investigators manually measured the traffic volume. Recently, with the development of information technology, research that automatically measures using images recorded with a digital video camera is carried out. However, the existing research [1] has a problem that it cannot count the people correctly at a congested situation in time resulting in many overlaps of people. There is a need for the establishment of a method to count the number of people correctly at the congested situation.

In this research, 3D information the moving region is calculated based on corresponding points [2] of the moving region with a histogram of brightness. We developed the system that can separate moving regions into each person by using cluster analysis [3] based on 3D information of the moving region and thus counts passing people based on the lap probability of a moving region.

The input data for this system are the animation and the background image of each digital video camera. We developed the system that implements the following functions: moving region extraction function; 3D measurement function; moving region separation function; and count passing people function. The output data of this system is the number of people who passed in each direction.

In this research, in order to indicate the validity of the method, we prove that we can count people correctly in a congested situation. The face of a stairway was filmed in stereo with two digital video cameras. The recognition rate as a result of this system experiment is 92% with an upward flow and 86% with a downward flow. From this result, the recognition rate in a downward flow is 7% lower than its counterpart in an upward flow. However, as a whole, more than 90% recognition rate was achieved. Therefore, it is demonstrated that this system can count correctly people in congested situation. However, in this system, the accuracy of the recognition rate decreases due to the speed of movement. Therefore, improving the measurement accuracy without associating with movement speed is the next issue.

Terada K., Ando T., Yamaguchi J., "Counting Passers-by Using a Color Camera", Journal of Robotics and Mechatronics, Fuji Technology Press, 11(2), 123-128, 1992.
Kitagawa E., Tanaka S., Furuta H., Sugimachi T., "A Research for Producing Web/3D Model House from 2D Digital Pictures", Journal of Applied Computing in Civil Engineering, Japan Society of Civil Engineers, 13, 295-302, 2004.
Anderberg R.M., "Cluster Analysis for Applications", Academic Press Inc, 1973.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description