emerging enforcement technology."
Biometrics: Facial Recognition
By: Susan Geoghegan
It is estimated that as many as
25 percent of complaint cases contain facial images of the suspects, which are
received from a variety of sources that include
social networks, cell phones, and other media. Improved system
architecture, increased accuracy of commercial algorithms, and inter-agency
standardization of the mug-shot capture process have greatly contributed to the
increased use of face recognition applications.
identification technologies to mobile applications allows officers in the field
to identify suspects while on foot, bicycle, or vehicle patrol. Available on
mobile phones and compact handheld devices, face recognition software is
generally user-friendly and has become much more affordable for use by
Officers on patrol that have
probable cause to detain or question an individual can ask for identification,
and if warranted, take a photo of the subject. The image is sent through a
secure Internet connection, checked against a database and, if there is a
match, information on the subject is quickly transmitted back.
How the Technology Works
While the history of computer-aided
facial recognition dates back to the 1960s, the first true success in facial
recognition technology occurred in the late 1980s and early 1990s with the
development of the eigenfaces technique. This technique is an appearance-based
approach to face recognition that uses a strong
combination of linear algebra and statistical analysis to generate a set of
basis faces—the eigenfaces—against which inputs are tested.
In 2006, the Face Recognition
Vendor Test, conducted by the National Institute of Standards and Technology,
revealed that computerized facial recognition technology had advanced
considerably, showing a tenfold improvement over comparable tests that had been
performed four years earlier. Since that time, the focus has been on improving
performance of face recognition technology on images captured under
The least intrusive of the
biometric methods, face recognition identifies a specific individual in a
digital image by analyzing and comparing patterns. In the past, facial
recognition software relied on 2D imaging for comparison or identification, and
was only effective when the images were taken in a controlled environment.
Even the smallest changes in
light or orientation could reduce the effectiveness of the system, as the
captured image could not be matched to any face in the database, resulting in a
high rate of failure. The emergence of 3D facial recognition technology greatly
improved the effectiveness of identification by capturing a real-time 3D image
of a person’s facial surface and using distinctive features of the face to
identify the subject.
With 3D face recognition
software, the system aligns and measures the captured image, and then
translates the facemap template into a unique code. If the database image is
3D, matching can occur without making any changes to the image. However, if the
database image is 2D, a complex algorithm is applied to convert the source
image into a 3D image in order to find a potential match.
For verification purposes, the
image is matched to only one image in the database. However, if the goal is to
identify a person, the source image is compared with several images available
in the database.
Trends in Face Recognition
According to most industry
experts, fingerprint biometric technology will continue to be the primary
modality used by law enforcement in the foreseeable future. Sean Mullin,
President and CEO of BI2 Technologies, attributes this trend to the high level
of accuracy offered by fingerprint biometrics, as well as the extensive
searchable database available to law enforcement jurisdictions nationwide.
However, he also believes that
facial and iris recognition will emerge as standard identification technologies
over the next decade. “The FBI’s Next-Generation Identification (NGI)
initiative is a clear indicator of the significant change that is quietly
taking place. [This] initiative includes the creation and implementation of
national iris and facial repositories—hosted
and maintained by the FBI—as a critical
addition to the improvements they are making to their identification
capabilities,” Mullin said.
John Hinmon, Senior VP of
Marketing for Cross Match Technologies, Inc., agrees that current searchable
databases for facial images are considerably smaller than those existing for
fingerprints. He also points out that facial algorithms still need refinement
with regard to coverings (glasses, etc.) and expressions (anger, stress,
But, like Mullin, Hinmon views
facial biometrics as an important tool for law enforcement. “As iris and
facial capture capabilities improve and algorithms improve, and the databases
continue to be populated, they will further solidify their place as ‘go to’
biometrics along with finger.”
President and CEO of
Animetrics, Inc., Paul Schuepp, sees law enforcement trending toward the use of
facial recognition systems as a complement to fingerprint technology. For
investigative processing, a captured latent facial photo can be used to filter
or search the mug-shot criminal database for possible matching candidates, but
it is not always conclusive evidence of suspect identification.
However, it may be the only
evidence to work with when investigators are unable to capture latent
fingerprints at the crime scene. “The trend is to now go deeper than before
because of facial technology on the investigations, especially when all you
have is video evidence,” Schuepp said. In addition, detectives are now able to
open up cold cases with the photographic facial image because of image
enhancement tools, such as Animetrics ForensicaGPS.
Robert Horton, Senior Director
of Marketing and Communications for MorphoTrak, also believes that fingerprint
will remain the dominant biometric in the government sector for many years to
come. Very few databases currently exist for iris technology, and face
recognition is too susceptible to issues with lighting, pose angle, and
“Iris has been captured by the
Department of Defense in Iraq
but few other databases exist, and today the vast majority of State and Local
police agencies do not capture iris,” Horton said. “The main issue with face is
lighting and pose angle (with different issues posed during both day and night
with shadows, contrast, wash-out, etc.).”
While Horton believes that advances
in face recognition technology will never be as accurate as fingerprint for
identification (a 1:N unknown person against a multi-million person database),
he does acknowledge that it is a useful tool when used in other applications.
Some examples include a check against 100 to 1,000 faces on a terrorist watch
list, a check against a banned person watch list at a casino, and a
verification check against a claimed identity (1:1).
A diverse range of facial
recognition solutions are available for mobile devices, all of which provide
accurate, reliable matching and rapid identification. Some systems are designed
for a specific identification modality, i.e., fingerprinting, iris scan, or face
recognition, while others combine two or more biometric technologies on one
device for in-the-field use.
By utilizing a combination of
biometric recognition techniques, the level of accuracy greatly increases. When
positive identification is not possible with one form of recognition technology
(fingerprint), a combination (fingerprint, iris, and face) can be used to
The Morpho RapID™, an
all-in-one handheld device offered by MorphoTrak, uses the most capable and
innovative technologies available on the market for on-the-spot ID checks,
immigration and border control, and suspect identification.
With full shift autonomy, this
mobile identification device is easy to use and features a rugged design that
conforms to IP65. Equipped with an on-board watch-list of up to 180,000
individuals, the Morpho RapID™ offers fingerprint and photo capture and real-time
high-quality WiFi and/or cellular wireless communication with central AFIS.
solution offered by MorphoTrak is the Morpho HIIDE5, a handheld military-grade
device that provides local or remote EBTS-compliant enrollment and search. It
is capable of generating onboard templates for fingerprint, iris and facial
images identical to those used in Morpho systems. With on-board databases that
hold up to 1.5 million records, HIIDE5 is ideal for high-volume remote applications.
The device feature a 5-inch color touchscreen, an on-screen keyboard for
entering demographic information, and is encryption and tamper-protected for
multi-biometric enrollment and identity management capabilities, the SEEK
Avenger from Cross Match Technologies, Inc. is a highly mobile and
interoperable system ideal for in-field operations. This rugged yet compact
device is built to perform in the challenging environments associated with
military, border security, and law enforcement deployments.
The device features
forensic-quality fingerprint and stand-off dual iris capture (even in direct
sunlight), high-resolution facial and evidence imaging, and multiple-format
credential reading. With optional 3G/4G wireless connectivity and an onboard
watch list of up to 250,000 records, the SEEK Avenger allows officers to
quickly access vital information, eliminating the need to transport the suspect
to the station house for identification.
BI2 Technologies’ MORIS™
(Mobile Offender Recognition and Identification System) is a handheld fully integrated
multi-modal biometric device for iris, fingerprint and facial recognition
capture. Named as one of the best innovations of 2010 by Popular Science, MORIS™ is designed to work with an iPhone. The app was developed by BI2 Technologies and utilizes
Animetrics’ facial recognition technology to automatically translate
two-dimensional images into three-dimensional pictures for enhanced analytical
After a digital image of an
individual’s fingerprint, iris or face is captured, it is transmitted over a
secure wireless network for comparison against an existing database of criminal
justice records. If a match is found, MORIS™ quickly provides identity
confirmation and previous criminal history. For facial capture, an officer
takes a photo of a person at a distance of about 2 feet to 5 feet and
approximately 130 distinguishing points on the face are identified, such as the
distance between a person’s eye and nose.
Mullen pointed out that MORIS’
relatively low cost ($3,000 including the Smartphone) makes it the ideal choice
for smaller law enforcement agencies. “MORIS™ enables the smaller agency to
have immediate access to our national iris biometric database for searches by
data (i.e., DOB, name, etc.) or by iris, as well as having the capability to
use the FBI certified fingerprint if their state’s Automated Fingerprint
Identification System (AFIS) can receive and process an FBI certified
fingerprint. MORIS™ can also perform facial recognition if their state has a
searchable mug shot repository.”
A leader in advanced 3D facial
recognition and facial identity solutions, Animetrics offers four patented
technologies for military and law enforcement applications. The FaceR™ MobileID
is a three-dimensional facial recognition and identity system that operates on
mobile devices, where it is managed by their
FaceR Identity Management Solution (FIMS) platform.
FIMS is a
comprehensive facial-recognition platform that provides centralized management
and advanced recognition services, including full search, matching and access
to multiple subject databases (including watch lists). User-friendly and easy to install, FaceR™
MobileID functions like other mobile
applications on Apple’s iOS for iPhone or the Android operating system.
This past May, Animetrics
announced the availability of ID-Ready, a subscription-based online service for
smaller police departments. The service applies 2D-to-3D algorithms to correct
a grainy, partial-view 2-D facial image, making it ready for most facial
recognition systems. Ideal for smaller agencies on limited budgets that cannot
invest in facial analytic systems, ID-Ready is offered in three pricing
options: pay as you go, monthly, and monthly prepaid. The company also extends
special pricing for qualified law enforcement agencies.
Police agencies interested in
utilizing facial recognition technology are advised by industry experts to
consider certain factors to ensure successful implementation, such as cost,
best practices, database access, and vendor support and training. Sean Mullin
suggests speaking with the potential vendor’s existing clients to assess the
performance of their technology, and confirm access to a national database.
“Make sure the vendor can
provide immediate and secure access to national biometric databases. Having
limited access to information from one jurisdiction or state significantly
reduces the value of the technology.”
Since there are significant
differences in cost, Mullin advises doing a price comparison with special
attention paid to hidden costs. Some companies include additional costs for the
number of enrollments or identity verifications, as well as click charges for
each time the agency checks for an individual.
“Our advice is to demand a
complete cost proposal—and only buy if
the vendor will provide: unlimited number of users; unlimited number of
enrollments; unlimited number of times authorized staff can check for an
individual in a national database; integration with existing systems and
databases (no redundant data entry).” On-site implementation and a minimum of
four hours of quality training during multiple shifts should also be included
in the cost, Mullen noted.
Robert Horton advises agencies
to follow the National Institute of Standards and Technology (NIST) best
practices in the capture of facial images, as there are specifications for
background color, lighting, maximum head tilt, etc. “Be sure to have the
software calculate the NIST-quality metrics for images, and re-capture when the
quality metric does not meet best practice guidelines,” Horton stated.
According to Paul Schuepp,
Animetrics offers training and programs for free evaluation use of the face
biometric applications to assist agencies in the decision-making process.
“Animetrics offers SaaS services online, such as http://id.ready.animetrics.com,
and for developers, Web service Facial Recognition API at http://api.animetrics.com.
The future trend will be to set up facial databases to share between law
Cross Match offers a wide range
of implementation services for its biometric-identity management solutions to
ensure smooth deployment. They have the experienced,
certified resources to serve on the front-end of solution deployments to
address the concerns around privacy, interoperability and adherence to
standards. From smooth initial implementation and timely integration to
post-deployment and day-to-day maintenance of the systems, Cross Match service
offerings can facilitate customers in any stage of biometric technologies
Susan Geoghegan is a freelance writer living in Naples, Fla. She can be
reached at firstname.lastname@example.org.