793 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
LL . . . . . . 
i" 
|: 
* * ■•¿•s.* * * 
’ ‘ “-S-S- “ “ * * 
1 i 
T~~ “ 
i“ 
I , 
» * « *» h» m «s m * 
(a) 
(b) 
Figure 9. Iterative solutions for the transformation parameters 
in (a). ” David_2 vs. David 1” and (b).” Ivanl vs. David 1”. 
For verification and identification purposes, acquired facial 
models were compared to available models. The comparison 
procedure required the co-registration of the facial models to a 
common reference frame and the matching of the registered 
models. The registration algorithm which combines the MIHT 
and ICPatch, was used to register and match the two facial 
models. Figure 7 and Figure 8 show the co-registration of the 
two facial models for a reference facial model Davidl. 
The scale factor of the transformation function was fixed in the 
experiments because the scale should not substantially differ 
between data acquisition epochs. In the experiment ” David_2 
vs. David 1”, the iterative solution (Figure 9(a)) for the 
transformation parameters revealed smooth and quick 
convergence. The estimated RMS of the normal distance 
between matching surface elements following the registration 
procedure was 0.629 mm. A large percentage of the points (Fig. 
7) were classified as matches (93.902%) with the non-matches 
mainly occurring around the edges of the facial models. The 
results showed a high quality of fit between two surfaces. In the 
experiment ” Ivan l vs. David l”, the iterative solution 
(Figure 9(b)) for the transformation parameters did not exhibit 
a smooth and rapid convergence. The RMS of the normal 
distances between the matched elements was 1.715 mm. The 
procedure achieved 79.458% of matched points (Fig. 8). 
Compared with ” David_2 vs. David l”, the results here 
showed a lower quality of fit between two surfaces. 
4. CONCLUSION 
This research presented a system for automated matching of 
facial models using a low-cost photogrammetric stereo system 
with pattern projection. The experimental results showed that, 
after calibration, low-cost digital cameras can reconstruct 3D 
facial models, and then facial model registration can be 
performed effectively. The proposed system has great potential 
for various applications such as surveillance, plastic surgery 
and personal verification. Because seven locations were 
sequentially imaged, movement of the subject may have 
produced errors. In the future, development of a system with 
multiple cameras, which can be controlled to capture images 
simultaneously, would improve both accuracy and matching 
without motion effect. Some limitations of the system are 
acknowledged. Certain surface details such as the hair and 
beard cannot be properly acquired by this system. 
REFERENCES 
Achermann, B., Jiang, X., and Bunke, H., 1997. Face 
recognition using range images, Proc. Int. Conf. on Virtual 
Systems and MultiMedia, pp. 129-136. 
Beumier, C., and Acheroy, M., 1995. Automatic Face 
Identification, Applications of Digital Image Processing XVIII, 
SPIE, vol. 2564, pp. 311-323. 
Chellappa, R., Wilson, C.L., and Sirohey S., 1995. Human and 
Machine Recognition of Faces: A Survey, Proceedings of the 
IEEE S3 (5), pp. 705-741. 
Cheng, R.W.T. and Habib, A., 2007. Stereo photogrammetry for 
generating and matching facial models. Optical Engineering 
46(6), pp. 67203-67213 
Jarvis, R., 1993. Range Sensing for Computer Vision, in: 
Three-Dimensional Object Recognition Systems, Jain, K. and 
Flynn, P.J., eds., Advances in Image Communication, vol 1, 
Elsevier Science Publisher, pp. 17-56. 
Habib, A., and Morgan, M, 2003. Automatic calibration of 
low-cost digital cameras. Optical Engineering, 42(4), 
pp.948-955. 
Habib, A., Pullivelli, A., Mitishita, E., Ghanma, M., and Kim E., 
2006. Stability analysis of low-cost digital cameras for aerial 
mapping using different geo-referencing techniques. Journal of 
Photogrammetric Record, 21(113), pp29-43. 
Harris C., and Stephens M.J., 1988. A combined comer and 
edge detector. A Ivey Vision Conference, pp. 147-152. 
Lao, S., Sumi, Y., Kawade, M., and Tomita, F., 2000. 3D 
template matching for pose invariant face recognition using 3D 
facial model built with iso-luminance line based stereo vision, 
Proc. ICPR, vol.2, pp.911-916. 
Schmid, C., Mohr, R. and Bauckhage C., 2000. Evaluation of 
interest point detectors. International Journal of Computer 
Vision, 37(2), pp. 151-172 
Yacoob, Y., and Davis, L.S., 1994. Labeling of human face 
components from range data, CVGIP: Image Understanding, 
60(2), pp.168-17