Partenaires et organisations internationales
(Anglais)
|
ACTISYSTEM, Dijon (F),
ART&MAGIC, Lieges (B),
BITS, London (UK), DURAN, Paris (F), EPFL - LIG, Swiss Federal Institute of Technology (CH)
|
Résumé des résultats (Abstract)
(Anglais)
|
TABSTRACT
In the framework of the project, we have proposed a motion capture methodology based on an anatomic human model. This model encompasses a precise anatomic description of the skeleton mobility associated with an approximated envelope. It has a double objective: by ensuring a high precision mechanical model for the performer, we can predict accurately the 3D location and the visibility of markers, thus reducing significantly the human intervention during the conversion
We introduced an approach to increasing the robustness of a motion capture system by using a sophisti-cated anatomic human model. It includes a precise descrip-tion of the skeleton's mobility and an approximated enve-lope.
It allows us to accurately predict the 3-D location and visibility of markers, thus significantly increasing the robustness of the marker tracking and assignment, and dras-tically reducing-or even eliminating-the need for human intervention during the 3-D reconstruction process. In con-trast to commercially available approaches to motion cap-ture such as the ones proposed by Elite tm and VICON tm , we do not treat 3-D marker reconstruction independently from motion recovery. Instead we combine these two pro-cesses and use prediction techniques to resolve ambiguities.
For example, we can predict whether or not a marker is ex-pected to be occluded by the body in one or more images and take this knowledge into account for reconstruction pur-poses.
When a marker cannot be reconstructed with cer-tainty from its image projections, we use the expected posi-tion of the skeleton to identify the marker and disambiguate its 3-D location. This is helpful when it is only seen by a small number of cameras. In our approach, the performer's skeleton motion is a byproduct of the reconstruction pro-cess.
Publications
M.C.Silaghi, R.Plaenkers, R.Boulic, P.Fua, D.Thalmann, Local and Global Skeleton Fitting Techniques for Optical Motion Capture, in: N. Magnenat Thalmann, D. Thalmann (eds), Modeling and Motion Capture Techniques for Virtual Environments, Lecture Notes in Artificial Intelligence, No1537, Springer, 1998, pp.26-40.
L. Herda, P. Fua, R. Plankers, R. Boulic, D. Thalmann, Skeleton-Based Motion Capture for Robust Reconstruction of Human Motion, Proc. Computer Animation 2000, IEEE CS Pre
|
