Human Head Pose Tracking In The Context Of Car Driving
Martin Breton
Denis Laurendeau (Supervisor)

This project's goal is to determine the head pose (orientation and position) of a person in the context of car driving. The human head being able to move from left to right, up and down, the system must be able to determine the orientation on about 180 degrees about all axes, or the equivalent of half a sphere.

Car driving is a task that implies a strong perceptual component. It is intuitive to think that there is a direct relationship between the deterioration of many visual capabilities and the possibility of having an accident.

In the case of our elders, even though they show a deterioration of many visual capabilities (acuity, greater glare and contrast sensibility), very few direct relationships have been established between these measurements and road accidents.

This project is part of a collaboration with the GRAME (Groupe de Recherche en Analyse du Mouvement et Ergonomie) and will be part of an instrumented simulator developed to measure the driver's behaviour in a controlled virtual environment.

A static and calibrated pair of cameras will be used as well as a target with markers in the form of a hat worn by the driver. The cameras will be placed on each side of the driver, facing him/her so that we can see the face from two different angles. This will allow the 3D position of the markers and thus the pose of the head to be obtained.
  • It can happen that part of the driver's face is not visible from one of the cameras. The system has to estimate a pose nonetheless.

  • The system must also be robust and precise.

The results will be used by the researchers of the GRAME in their Car driving and getting older study and is part of COBVIS: A simulation environment for the evaluation of cephalo-ocular behaviour and visuel pattern on drivers, a project from Auto21.
Expected results:

The head pose is expected to be found for each video frame of the sequence.
Calendar: September 2003 - September 2005
Last modification: Sep 28 2007 2:41PM by mbreton


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