MONNET Simulation
Richard Drouin
Denis Laurendeau (Supervisor)
Problem: Tracking people in complex environments is an important problem in several areas : security in public places, military interventions in urban areas or in assisting people experiencing a loss of autonomy. Our project is part of the MONNET project which aims to develop a system for tracking people. This system is composed of an ensemble of nodes each equipped with visible and infrared cameras as well as a calculation unit for the treatment of video sequences. These nodes are connected together in a network and exchange information on people who are moving within their field of view. The network is composed of these nodes which are loosely coupled since each node is an independent component of the network, but is able to exchange information with the other nodes so as to enable a more complete tracking of people. Moreover, the fields of view of the nodes are not necessarily overlapping, which implies that the system must be able to treat information while taking into account the non-observed zones of the environment. Our project consists in studying the aspects related to the communication… between loosely coupled nodes and to define an exchange protocol enabling the nodes to exchange relevant information regarding the static and dynamic appearance of people who are moving within their field of view.
Motivation: In the literature most visual systems enabling the tracking of people are described as being based on a central server which is responsible for managing the interactions between nodes. The system described herein is significantly different from these systems since it is based on a network of loosely coupled cameras.
Approach: The research methodology consists in carrying out a literature survey on comparable systems. We will then perform an analysis of the needs relative to the desired application and define a software architecture responding to these needs. A simulation environment with different possible solutions will be implemented on the APIA simulator presently available in the Laboratory. The implementation of this architecture will be carried out using the C++ language and the prototype will be tested with simulated data as well as data provided by the algorithms of artificial vision developed by other members of the research team.
Challenges: The main challenge of this project consists in developing a flexible, robust and evolving inter-node communication module.
Applications: The results of our work will be integrated into the research involving the MONNET system.
Last modification: Jun 11 2004 10:06AM by drouin04


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