The tracking and detection of people by means of a computerized system of cameras has been the subject of many research projects in the recent time. These systems use different types of models to recognize people in order to track them in a monitored area. Gait modeling is considered in these systems since it is possible to recognize somebody by observing his walk at a distance, and without his cooperation.

Most of the current methods for gait modeling assume that one will be able to observe people from a side view, which is obviously not always the case. Gait modeling for any view is then problematic since the motion of the body parts are distorted by the perspective projection effect. A method is thus proposed in order to obtain normalized body part trajectories for a walking person. This means that body part trajectories are not dependent on the view from which the person is observed, and that the trajectories seem to be observed from a side view. The proposed method does not necessitate a priori knowledge of the scene or the intrinsic and extrinsic camera parameters. A method is also proposed in order to evaluate the normalized trajectories.

The CVSL seminars are held on Fridays at 11:30 a.m. in room PLT-2700.