A new technique for 3D human posture estimation is proposed. The technique relies on the use of two orthogonal cameras in a controlled scenario. The actor region is segmented in both views and the 3D posture is estimated from the detection of the five crucial points (head, hands and feet) in both images. Crucial points candidates are defined as the local maxima of the geodesic distance with respect to the center of gravity of the actor region analyzed in the boundary of the actor region (silhouette). Selected crucial points are classified as head, hands or feet using a set of geodesic distances computed on a robust morphological skeleton of the actor region. Results in both views are fused to obtain the 3D posture estimation, while allowing for solving some possible inconsistencies of the classification process. Further robustness is introduced in the system by tracking the crucial point positions in time.
Cite as: Correa, P., Marques, F., Marichal, X., Macq, B. (2004) 3d human posture estimation using geodesic distance maps. Proc. 9th Conference on Speech and Computer (SPECOM 2004), 31-34
@inproceedings{correa04_specom,
author={Pedro Correa and Ferran Marques and Xavier Marichal and Benoit Macq},
title={{3d human posture estimation using geodesic distance maps}},
year=2004,
booktitle={Proc. 9th Conference on Speech and Computer (SPECOM 2004)},
pages={31--34}
}