Short Paper Code:
O07
Title:
Non Conventional 3D Human Face Verification System
Authors:
Zimmermann, A.C; acz@labmetro.ufsc.br;
Encinas, L.S.; lse@labmetro.ufsc.br
Gonçalves Jr., A. A.; aag@labmetro.ufsc.br
Barreto,
J.M.; barreto@inf.ufsc.br
LABMETRO
Federal University of Santa Catarina
Brazil (88040970), Florianopolis, SC
ABSTRACT
Different
from traditional methods which use two-dimensional images and gray levels to
recognize human faces, this article shows a known shape extraction methodology applied to the
extraction of 3D human faces conjugated with a conventional and non conventional
algorithms for face verification. The SORFACE project involves two main knowledge areas, 3D
shape extraction and pattern recognition. The first is based on Fourier Profilometry and the second on Case Base Reasoning - CBR and Artificial Neural Networks - ANN, which perform a
symbolic and connectionist recognition system. Although these methodologies
themselves are not new, the goal of this work is
conjugate all in a face verification application problem and shows the results. Are commented too the benefits achieved
by this 3D extraction technique over the illumination and geometric positioning
problems. This is only viable today thanks to the increase of processing
capacity of the new computers. This article describes all the techniques used
to build a non conventional optical system for 3D human face verification, the SORFACE project
proposal.