D'Apuzzo, Nicola
3 USE OF 3-D DATA FOR HUMAN BODY MODELING
From the multi-image sequence, the process described extracts
data in form of a 3-D point cloud of the visible body surface at
each time step and a vector field of 3-D trajectories. Figure 16
shows the results achieved by a 2-D contour tracking algorithm Figure 16, Results of the silhouette tracking process
using the 3-D trajectories. The algorithm is based on the snake &
technique (Kass et al. 1988). Given an extracted contour in one
frame, the trajectory information of surrounding 3-D points,
projected onto the image plane, is used to predict the position of
the contour in the next frame. The silhouette information and the
measured 3-D points for each frame are used to fit a complete
animation model to the data. The results of the fitting process are
shown in Figure 17. For the detailed explanation of the process
we refer to the related publication (Plaenkers et al. 1999).
4 CONCLUSIONS AND FUTURE WORK Figure 17. Results of the fitting process
A process for an automated extraction of 3-D data from multi-image sequences has been presented. The extracted 3-D
data is composed of two parts: measurement of the body surface at each time step of the sequence and a vector field of
3-D trajectories (position, velocity and acceleration). Initially, the two different types of data are very noisy, therefore
adequate filters have been developed and applied to the data.
Lot of work still remains for the future to improve the quality of the extracted 3-D data. For the surface measurement, the
most important feature which has to be integrated in the process, is the definition of geometric and neighborhood
constraints in the least squares matching algorithm. The consideration of neighborhood information should be also
integrated in the tracking process to achieve more reliable results.
In addition, the gain in robustness and level of automation should be also considered, since the final goal of the project is
the development of a fully automated and robust process.
ACKNOWLEDGEMENTS
The work reported here was funded in part by the Swiss National Science Foundation.
REFERENCES
Anai T., Chikatsu H., 1999, Development of Human Motion Anal
ysis and Visualisation System. Int. Archives of Photogrammetry and
Remote Sensing, Onuma, Japan, 32 (5-3W12), pp. 141-144
Badler N. I., 2000. Animation 2000++. IEEE Computer Graphics and Applications, 20(1), pp. 28-29
Badler N. I., Palmer O., Bindiganavale R., 1999. Animation Control for Real Time Virtual Humans. Comm. ACM, 42 (8), pp. 65-73
Bhatia G., Smith K. E., Commean P. K., Whitestone J ., Vannier M. W., 1994. Desi
gn of a Multisensor Optical Surface Scanner. Sensor
Fusion VII, SPIE Proc., Boston, USA, Vol. 2355, pp. 262-273
Boulic R., Fua P., Herda L., Silaghi M., Monzani J .-S., Nedel L., Thalmann D., 1998. An An
atomic Human Body for Motion Capture.
Proc. EMMSEC”98, Bordeaux, France
Boulic R., Thalmann D., Molet T., Huang
Z., 1997. An animation interface designed for motion capture. Computer Animation ‘97
Conference, IEEE Press, pp. 77-85
Bradtmiller B., Gross M., 1999, 3D Whole Body Scans: Measurement Extraction Software Validation. “
Digital Human Modeling for
Design and Engineering" Conference and Exposition, May 1999, Hague, The Netherlands
Brown D. C., 1971. Close-Range Camera Calibration. Photogrammetric Engineering and Remote Sensing, 37 (8), pp. 855-866,
Brunsman M. A., Daanen H., Robinette K. M., 1997. Optimal Postures and Positioning
International Conference on Recent Advances in 3-D Digital Imaging and Modelin
CA, USA
for Human Body Scanning. Proc. of
g, IEEE Computer Society Press, Los Alamitos,
Burnsides D. B., 1997. Software for Visualization, Analysis and Manipulation of Laser Scan Images.
Paper #3023-15, ASC-97-0378,
Internal Report Armstrong Laboratory, Wright-Patterson Air Force Base, Ohio, USA
170 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000.
Cer!
on 2
Cor
pp.
Da:
of I
D’/
Rer
Dy:
htt
Fu:
seq
Gn
Re
Ga
Vis
Ph
Joi
5%
CS ITA AN AAA CU et