Share Email Print

Proceedings Paper

High-accuracy and robust face recognition system based on optical parallel correlator using a temporal image sequence
Author(s): Eriko Watanabe; Mami Ishikawa; Maiko Ohta; Kashiko Kodate
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Face recognition is used in a wide range of security systems, such as monitoring credit card use, searching for individuals with street cameras via Internet and maintaining immigration control. There are still many technical subjects under study. For instance, the number of images that can be stored is limited under the current system, and the rate of recognition must be improved to account for photo shots taken at different angles under various conditions. We implemented a fully automatic Fast Face Recognition Optical Correlator (FARCO) system by using a 1000 frame/s optical parallel correlator designed and assembled by us. Operational speed for the 1: N (i.e. matching a pair of images among N, where N refers to the number of images in the database) identification experiment (4000 face images) amounts to less than 1.5 seconds, including the pre/post processing. From trial 1: N identification experiments using FARCO, we acquired low error rates of 2.6% False Reject Rate and 1.3% False Accept Rate. By making the most of the high-speed data-processing capability of this system, much more robustness can be achieved for various recognition conditions when large-category data are registered for a single person. We propose a face recognition algorithm for the FARCO while employing a temporal image sequence of moving images. Applying this algorithm to a natural posture, a two times higher recognition rate scored compared with our conventional system. The system has high potential for future use in a variety of purposes such as search for criminal suspects by use of street and airport video cameras, registration of babies at hospitals or handling of an immeasurable number of images in a database.

Paper Details

Date Published: 16 September 2005
PDF: 8 pages
Proc. SPIE 5908, Optical Information Systems III, 590815 (16 September 2005); doi: 10.1117/12.616551
Show Author Affiliations
Eriko Watanabe, Japan Women's Univ. (Japan)
Mami Ishikawa, Japan Women's Univ. (Japan)
Maiko Ohta, Japan Women's Univ. (Japan)
Kashiko Kodate, Japan Women's Univ. (Japan)

Published in SPIE Proceedings Vol. 5908:
Optical Information Systems III
Bahram Javidi; Demetri Psaltis, Editor(s)

© SPIE. Terms of Use
Back to Top