Nowadays face recognition systems have many application fields. Unfortunately, lighting variations and ageing effects are still open issues. Moreover, face changes over time due to ageing. A further problem is due to occlusions, for example the glass presence. Re-enrolling user’s face is time-consuming and does not solve above problems. Therefore, unsupervised template update has been proposed, and named self update. Basically, this algorithm adapts/modifies templates or face models by collecting samples during system operations. The most effective variant of self update is based on the collection of multiple templates. However, this approach has been evaluated and tested in conditions under which the possible number of collectable templates is uncostrained. Actually, available resources are limited in memory and computational power, thus it is likely that it is not possible to have more than a pre-set number of templates. In this paper, we propose a classification-selection approach, based on the combination of self update and C-means algorithms, which keeps constant the number of templates and improve the ratio between intra-class variations and inter-class variations for each user. Experimental results show the effectiveness of this method with respect to standard self update.

A classification-selection approach for self updating of face verification systems under stringent storage and computational requirements

TUVERI, PIERLUIGI;MURA, VALERIO;MARCIALIS, GIAN LUCA;ROLI, FABIO
2015-01-01

Abstract

Nowadays face recognition systems have many application fields. Unfortunately, lighting variations and ageing effects are still open issues. Moreover, face changes over time due to ageing. A further problem is due to occlusions, for example the glass presence. Re-enrolling user’s face is time-consuming and does not solve above problems. Therefore, unsupervised template update has been proposed, and named self update. Basically, this algorithm adapts/modifies templates or face models by collecting samples during system operations. The most effective variant of self update is based on the collection of multiple templates. However, this approach has been evaluated and tested in conditions under which the possible number of collectable templates is uncostrained. Actually, available resources are limited in memory and computational power, thus it is likely that it is not possible to have more than a pre-set number of templates. In this paper, we propose a classification-selection approach, based on the combination of self update and C-means algorithms, which keeps constant the number of templates and improve the ratio between intra-class variations and inter-class variations for each user. Experimental results show the effectiveness of this method with respect to standard self update.
2015
9783319232331
9783319232331
Computer Science (all); Theoretical Computer Science
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11584/177496
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 4
social impact