Face Recognition Using Video Clips and Mug Shots
Gaile G. Gordon
Marquess E. Lewis
TASC, Inc.
55 Walkers Brook Drive, Reading, MA 01867
Abstract
This paper presents two automated identification systems
based on 3D face recognition developed at TASC. They have the
capability to processes mug shots, consisting of frontal and profile
views, as well as video. Both systems are discussed in terms of their
advantage with respect to viewing angle changes. The mug shot system,
FaceMatcher, is based on low level pattern recognition of
geometrically normalized subimages. The addition of profile imagery is
shown to dramatically lower the recognition error rates of a frontal
view based system. The video based system, FaceMatcher3D, explicitly
extracts the 3D pose of the head in each video frame. Low level
features, such as corners, are identified and tracked in the video
stream. The feature tracks are processed by a shape from motion
algorithm which produces estimates of 3D geometry and pose. The
geometry and pose estimates are considered together with facial
structure constraints, temporal constraints, and initial pose
estimates to refine knowledge of the specific face structure and its
pose. This information can be used to extract frontal and profile
views from the sequence and then to geometrically normalize them for
comparison with the database. The capability to perform face
recognition from video presents important advantages over other
biometric identification methods, such as finger prints or retinal
scans, because identification is now passive. Passive identification
can be used not only in traditional access control, but also in
surveillance applications.
G. Gordon, M. Lewis, ``Face Recognition Using Video Clips and Mug Shots'',
Proceedings of the Office of National Drug Control Policy (ONDCP) International
Technical Symposium (Nashua, NH), October 1995.
Paper reprinted by permission of TASC, Inc., who holds the copyright thereto.