Since: Dec 13, 2005
Last update: Aug 27, 2009

Computer Vision for Network Security

Surveillance Camera System for Mobile Computing Environments using an Active Zooming Camera and MAC Address Tracking

Hideaki Goto
Information Synergy Center, Tohoku University, Japan

Proceedings Image and Vision Computing New Zealand 2005 (IVCNZ 2005),
pp.108-113, 2005 (Nov. 28-29, Dunedin, New Zealand).
(ISBN 0-473-10523-3)

Paper in PDF (2.1MB)

Abstract

Mobile computing environments using wireless networks and Ethernet jacks have become popular recently in universities, cafes, on the streets, etc. Although many of the network access systems are protected by user authentication, they are sometimes attacked by people who try to use the network illegally. It is difficult for network administrators to locate, track, and identify users in such a mobile environment because people travel from site to site. If a conventional, fixed surveillance camera is used for monitoring a big field, it cannot take pictures whose resolution is high enough for personal identification. To deal with these problems, we propose an intelligent surveillance camera system that utilizes an active camera for tracking and zooming into network users for better personal identification. The system is also equipped with a MAC address tracking mechanism that is useful for reducing video data to be recorded.

keywords:   surveillance camera, wireless network, active camera, user tracking, MAC address tracking

Demo

Please see the paper for sample images.
Here is a demo with a movie.

A test scene in our (messy) laboratory. There are three moving objects instead of humans in the scene.

An image sequence taken by the active camera.

• Movie in AVI (Cinepak Codec, 24fps, 63.9MB)
  This AVI file can be viewed by various players including xanim.
  The original image sequence was captured at 30fps via NTSC-DV converter.

Description of the events :

• [0:00]   It's dim in the room at the beginning. Target switching occurs every 3 seconds.
• [0:15]   The lights in the room have been turned on. The system becomes unstable for a while until it finishes learning the background.
• [0:34]   The background learning has finished. The system resumes the normal operation.
• [0:50]   A stuffed chameleon is put in the scene.
• [0:52]   The chameleon is focused, because it is recognized as a foreground object at first.
• [1:05]   We don't see the chameleon in the second round of the target switching, because it has been incorporated into the background.
• [1:11]   The chameleon has been moved a bit.
• [1:14]   We see the chameleon again (as a moving object).
• [1:27]   We don't see the chameleon.