Inspired by Bertillon - Recognition Based on Anatomical Features from 3D Face Scans

Czech title

Rozpoznávání obličejů založené na anatomických rysech

Type

conference paper

Language

english

Authors

Mráček Štěpán, Ing. (DITS FIT BUT)
Busch Christoph, prof. Dr. (FhG-IGD)
Dvořák Radim, Ing. (DITS FIT BUT)
Drahanský Martin, prof. Ing., Dipl.-Ing., Ph.D. (DITS FIT BUT)

Keywords

Biometry, Anthropometry, Face Recognition, FRGC, 3D Face

Abstract

We present an automatic 3D face recognition algorithm that is inspired by Alphonse Bertillon's anthropometry. Our recognition pipeline consists of several steps. First, the facial landmarks such as the tip of the nose or the inner eye corners are detected. Subsequently the head rotation is compensated during the orientation normalization process. The facial features are extracted by performing 61 different measures. We also present a feature evaluation function that rates individual components of the feature vector. Finally, our results are compared with two other 3D face recognition methods. We show that the multi-algorithmic system consisting of the anatomical-based recognition together with the eigenfaces method and the recognition using histogram-based features reaches significantly better results than any of the employed methods individually.

Annotation

We present an automatic 3D face recognition algorithm that is inspired by Alphonse Bertillon's anthropometry. Our recognition pipeline consists of several steps. First, the facial landmarks such as the tip of the nose or the inner eye corners are detected. Subsequently the head rotation is compensated during the orientation normalization process. The facial features are extracted by performing 61 different measures. We also present a feature evaluation function that rates individual components of the feature vector. Finally, our results are compared with two other 3D face recognition methods. We show that the multi-algorithmic system consisting of the anatomical-based recognition together with the eigenfaces method and the recognition using histogram-based features reaches significantly better results than any of the employed methods individually.

Published

2011

Pages

53-58

Proceedings

Proceedings of the 3rd International Workshop on Security and Communication Networks

Conference

3rd International Workshop on Security and Communication Networks, Gjovik, NO

ISBN

978-82-91313-67-2

Publisher

Gjovik University College

Place

Gjovik, NO

DOI

10.1109/IWSCN.2011.6827717

UT WoS

000397201800008

EID Scopus

2-s2.0-84903581562

BibTeX

@INPROCEEDINGS{FITPUB9696,
   author = "\v{S}t\v{e}p\'{a}n Mr\'{a}\v{c}ek and Christoph Busch and Radim Dvo\v{r}\'{a}k and Martin Drahansk\'{y}",
   title = "Inspired by Bertillon - Recognition Based on Anatomical Features from 3D Face Scans",
   pages = "53--58",
   booktitle = "Proceedings of the 3rd International Workshop on Security and Communication Networks",
   year = 2011,
   location = "Gjovik, NO",
   publisher = "Gjovik University College",
   ISBN = "978-82-91313-67-2",
   doi = "10.1109/IWSCN.2011.6827717",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/9696"
}