MONNET: Monitoring of Extended Premises: Tracking Pedestrians Using a Network of Loosely Coupled Cameras (Precarn Inc - PUL Program)

The MONNET research initiative aims at developing an intelligent computer vision-based monitoring system for non-intrusive and real-time automatic tracking of persons moving around in public premises. Tracking a person is an important problem that arises in many monitoring or surveillance applications. It is required, for instance, in order to detect and identify suspicious behaviour, or, in a more constructive setting, to guide a person towards a given destination. Such an active monitoring system based on passive sensing (e.g. using ambient lighting sources) has a wide diversity of potential applications, not limited to security issues. For instance, the monitoring system mentioned above might be used to track people in a retirement home and to detect unusual behaviour such as falling down or being hit by a seizure. In addition, a computer vision-based system would make it possible to create databases memorizing and summarizing events and behaviour relevant to future monitoring contexts. These databases could be shared between sites in a network and relied on in globally improving their efficiency.

More specifically, the proposed computer vision-based monitoring system aims to analyze image sequences from a network of cameras disposed at various locations sparsely covering the extended premises under observation and connected to computing "nodes". Each node of the system needs to detect and track people in a field of view of its assigned cameras, characterize these people in terms of their shape and appearance, memorize the time interval during which the people are visible, estimate the direction taken by these people, broadcast relevant information to other nodes on the network, and build a log file describing the activity that has occurred in the monitored areas.

Project title    Investigator(s)    Level
   Modelling and Comparing Human Gait from Monocular Video Sequences    Frédéric Jean (Student)
Robert Bergevin (Supervisor)
Alexandra Branzan-Albu (Co-supervisor)
(5/2005 - 5/2009)
   Incremental description of deformable objects in video sequences using computer vision    Stéphane Drouin (Student)
Marc Parizeau (Supervisor)
Patrick Hébert (Co-supervisor)
(1/2003 - 12/2007)
   Fusion of infrared and visible images for the tracking of pedestrians    Vincent Grégoire (Student)
Xavier Maldague (Supervisor)
Denis Laurendeau (Co-supervisor)
(1/2004 - 4/2005)
   Low-level segmentation and tracking of people with a network of independent cameras    Nicolas Martel-Brisson (Student)
André Zaccarin (Supervisor)
(1/2003 - 1/2005)



©2002-. Computer Vision and Systems Laboratory. All rights reserved