Pattern Recognition Letters
Special Issue on
Novel Pattern Recognition-Based Methods
for Reidentification in Biometric Context
CALL FOR PAPERS
http://www.face-rec.org/biocon/
The object reidentification problem is to authenticate an object
across multiple disjoint fields of view. Towards that end, once
the object has been detected and initialized at one location, one
seeks to match it with a feasible set of candidates detected at
other locations and over time. Standard pattern recognition
methods can be exploited, including Bayesian analysis, in order
to cope with the dynamic nature of the problem. In the context of
biometrics, advanced video (data stream) surveillance
applications (including sensor networks) for airport and subway
surveillance require the ability to reidentify an individual.
Face- and gait- based recognition are especially promising
biometrics for reindentification, since they can operate at a
distance and do not require a detailed and/or high resolution
image of the subject and/or its biometric traits. Two core
aspects this special issue will address are: (1) registration to
establish correspondences between parts of a pair of images; and
(2) invariant template representations suitable to compare the
corresponding parts. The invariance requirement comes from the
inherent variability in the data capture process with respect to
both sensors and subjects, e.g., PIE (pose, illumination, and
expression).
The methods proposed are expected to take advantage among others
of (a) recognition and tracking are complementary to each other;
(b) temporal reasoning and spatial layout of the different
cameras can be used for pruning the set of candidate matches; and
(c) the brightness transfer function between different cameras
can be used to track individuals over multiple non-overlapping
cameras. Further help to handle image variability comes from on-
line evidence accumulation characteristic of closed –loop
control, e.g., explore and exploit using sequential importance
sampling (SIS). Learning and adaptation using both labeled and
unlabeled data using statistical learning, in general, and semi-
supervised learning provide further help with reidentification.
One particular earning strategy of interest is co-training, where
only a small amount of labeled data is required to learn and the
use of unlabeled data improves performance over time.
The solution to the reidentification problem would make a
significant contribution to wide range of biometric applications.
Reidentification (reverification) of user’s identity is of
paramount importance for high security applications, when single
and static verification might not ensure that unauthorized users
do not gain control of a communication channel (e.g. a mobile
connection) during a critical service. The methods proposed
should be as unobtrusive as possible and require minimal
interaction from the user. Reidentification has also the
potential to foster dynamic multibiometric environments.
Currently, multibiometric systems usually implement static
architectures, which do not change over time. On the other hand,
the progressive addition of new modules would contribute to
better recognition performance. The addition of a new module to a
fully operational system would usually require its preliminary
setup and training. This involves a full enrolling and labelling
phase. A robust surveillance strategy that has access to and
exploits reidentification could mostly operate in an unsupervised
fashion and without explicit annotation System – like methods are
especially encouraged.
We invite original contributions that provide novel solutions to
challenging problems. Submitted papers can address theoretical or
practical aspects of progress and directions in biometrics
research.
Topics of interest include, but are not limited to:
o Face reidentification in video surveillance and sensor networks
o Gait-based reidentification
o Latent ant live fingerprint reidentification
o On-line learning from data streams for biometric reidentification
o Automatic annotation / labeling of biometric templates
o Biometric identity management
o Co-training and reidentification
o Metrics, protocols, and performance evaluation
o Multibiometrics, data fusion, and reidentification
o Reidentification for high security applications
o Reidentification in dynamical systems
SUBMISSION OF MANUSCRIPTS
Authors are encouraged to submit their papers electronically by
using online manuscript submission at:
http://ees.elsevier.com/patrec/
To ensure that all manuscripts are correctly identified for
inclusion into the special issue, it is important that authors
select the acronym “BIOCON” of this special issue when they reach
the “Article Type” step in the submission process.
All papers will be rigorously refereed and will undergo a very
competitive selection process. The length of your manuscript
should not exceed 7000 words, plus the necessary figures and
tables.
SUBMISSION PERIOD
The Elsevier Editorial System (http://ees.elsevier.com/patrec/)
will be set in due time to allow authors to upload their
contributions to the special issue in the period June 15, 2011 –
July 15, 2011.
IMPORTANT DATES
Manuscript submission deadline: *** July 15, 2011 extended deadline ***
First notification: September 15, 2011 *
Revised manuscript submission: November 15, 2011 *
Notification of final decision: November 30, 2011 *
Publication of the special issue: 2012 *
* tentative dates
For additional information, please contact one
of the GUEST EDITORS at the addresses below:
Mislav Grgic
University of Zagreb
Zagreb, Croatia
www.vcl.fer.hr/mgrgic
Michele Nappi
University of Salerno
Fisciano (SA) 84084, Italy
mnappi@unisa.it
Harry Wechsler
George Mason University
Fairfax, VA 22030, USA
wechsler@gmu.edu