Abstract
This paper presents an approach to build a 3D human-body model from uncali
brated monocular video sequences. Using an orthographic projection model for
the camera and a generic human body model, we constitute a 2D-3D conver
sion scheme to generate a 3D skeletal model from the 2D image domain. The
complete 3D geometry and kinematics information of a human being can be ex
tracted without any a-priori information about the camera system or the person
involved. Our promising experimental results can accurately reconstruct the hu
man motion trajectory. The system can be utilized for an effective human-motion
capture/analysis and a 3D skeleton reconstruction of deformable objects.
Original language | English |
---|---|
Title of host publication | Proceedings of the 27th Symposium on Information Theory in the Benelux, 8-9 June 2006, Noordwijk, The Netherlands |
Place of Publication | Delft |
Publisher | Werkgemeenschap voor Informatie- en Communicatietheorie (WIC) |
Pages | 299-306 |
ISBN (Print) | 978-90-71048-22-7 |
Publication status | Published - 2006 |