Combined thermal and colour sensors for enhanced detection and tracking in an uncontrolled environment
Louis St-Laurent
Xavier Maldague (Supervisor)
Problem: Most video surveillance systems currently available on the market use only the visible range of the electromagnetic spectrum. Moreover, the solutions proposed to date for automatic tracking of moving objects in a scene are often too computationally expensive to be carried out in real time and their performance deteriorates significantly in uncontrolled environments (such as outdoors).
Approach: The objective of this thesis is to exploit a combination of thermal and visible data to render the detection and tracking of objects of interest in uncontrolled environments more efficient. Within this project, one single fixed acquisition node will be considered. A node is composed of a thermal camera sensitive to wavelengths between 8 and 12 micrometres (LWIR) and a colour visible camera.
Challenges: Four challenges will be addressed:
  • alignment of thermal and visible images for a large depth of field;
  • real time segmentation of moving objects in an uncontrolled environment;
  • tracking of multiple objects in the presence of occlusions, regroupings and separations;
  • adaptive combination of thermal and visible data.
Special attention will be devoted to the rapidity of execution so as to attain a processing speed of 10 frames per second.
Calendar: May 2005 -
Last modification: 2007/11/09 by lstlaure


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