Autori: Nicolaie Popescu-Bodorin, Valentina E. Balas

Editorial: J. Fodor, R. Klempous, C.P. Suárez Arauj, Springer Verlag, Recent Advances in Intelligent Engineering Systems, p.405-430, 2011.

Rezumat:

This chapter discusses the latest trends in the field of evolutionary approaches
to iris recognition, approaches which are compatible with the task of
multi-enrollment in a biometric authentication system based on iris recognition, and
which are also able to ensure strong discrimination between the enrolled users. A
new authentication system based on supervised learning of iris biometric identities
is proposed here. It is the first neural-evolutionary approach to iris authentication
that proves an outstanding power of discrimination between the intra- and interclass
comparisons performed for the test database (Bath Iris Image Database). It is
shown here that when using digital identities evolved by a logical and intelligent
artificial agent (Intelligent Iris Verifier/Identifier) the separation between inter- and
intra-class scores is so good that it ensures absolute safety for a very large percent
of accepts (97%, for example), i.e. recognition is no longer a statistical event, or in
other words, the statistical aspect of iris recognition becomes residual while the logical
binary aspect prevails. In this way, iris recognition theory and practice advance
from inconsistent verification to consistent verification/identification.

Cuvinte cheie: neural-evolutionary approaches to iris recognition // neural-evolutionary approaches to iris recognition

URL: http://fmi.spiruharet.ro/bodorin/articles/2011-sv-bodorin-balas.pdf