2015 | OriginalPaper | Buchkapitel
Nonrigid Point Set Registration-Based 3-D Human Pose Tracking from Depth Data
verfasst von : Dong-Luong Dinh, Nguyen Duc Thang, Sungyoung Lee, Tae-Seong Kim
Erschienen in: 5th International Conference on Biomedical Engineering in Vietnam
Aktivieren Sie unsere intelligente Suche um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
In this paper, we present a novel approach of recovering a 3-D human pose from a single human body depth silhouette using nonrigid point set registration. In our methodology, a human body depth silhouette is presented as a 3-D points set that is matched to the next 3-D points set through point correspondences between them. To recognize and maintain the body part labels, we first initialize the initial points set and their corresponding body parts, then transform them to the next points set according the point correspondences via nonrigid point set registration. Upon the point registration, we use the information of the transformed body labels of the registered pose to create a human skeleton model. Finally, a 3-D human pose is recovered by mapping the skeleton’s position and orientation information to a 3-D synthetic human model. Our quantitative and qualitative evaluation on synthetic and real data show that complex poses could be tracked and recovered reliably.