| |
Home
|
|
 |
Messages
|
|
| |
|
|
| |
Members
Only |
|
| |
Chat |
|
| |
Files |
|
| |
Photos |
|
| |
Links |
|
| |
Database |
|
| |
Members |
|
| |
Calendar |
|
| |
Promote |
| | |
|
From:  "James Northstar" <http://groups.yahoo.com/group/phantomtruth/post?protectID=091233091165042233050098109252176090177098100009128121188150166091061>
Date:
Wed Jun 4, 2003 11:58 pm
Subject: DIVA: the eye in the
sky
|
ADVERTISEMENT
 | |
Distributed
digital video arrays, or DIVAs, are collections of really
smart
cameras able to detect and identify an individual in a crowded
train station and track him wherever he goes -- out of the
station,
into the parking lot, onto the freeway and so on.
They also notify authorities when they "think" the
individual engages
in suspicious activity or meets with
questionable cohorts.
You can watch for these DIVAs in
summer 2004.
The Department of Defense awarded $600,000 to
University of
California at San Diego's Computer Vision and
Robotics Research lab
Friday for further development of DIVAs,
cameras that see, think and
communicate.
Granted through
a federal counterterrorism interagency task force
called the
Technical Support Working Group, the funding is slated to
help
CVRR redirect these intelligent camera systems from their
initial intent, which was preventing traffic delays, to stopping
terrorism.
For the past four years, CVRR's DIVAs
assessed traffic patterns,
located accidents and notified
firefighters of emergencies, according
to Mohan Trivedi,
director of the DIVA project and professor at
UCSD's Jacobs
School of Engineering. This year, the DIVA technology
provided
extra security at the Super Bowl, both around the stadium
and in
San Diego's Gaslamp district.
"The local police wanted to
make sure the crowds weren't unmanageably
large and rowdy,"
Trivedi said. "Our integrated system (analyzed)
crowd sizes, not
individual people. If the number of people exceeded
a certain
limit, notification would be sent to authorities."
Trivedi's
research and development focus is shifting from crowd
control to
what he calls "personal security" for important locations.
"Instead of having guards for 24 hours consistently,"
Trivedi
said, "we have the DIVA architecture that can
immediately detect and
provide hi-res video of certain events."
Even with the advent of the Department of Homeland Security
and
orange security alerts, round-the-clock guards have never
been
feasible for every national landmark or "important
location." But
cameras, installed and mobile, just may be.
On a sunlit pier, moonlit corner or crowded sidewalk, or in
a
deserted back alley, DIVA systems can observe an individual or
group,
anticipate behavior and trigger complex chains of
camera-to-camera
communication if the system determines that
someone looks, moves or
behaves in a certain way.
The
capability to identify a man automatically based on his facial
structure, or to locate a woman digitally based on her
distinctive
gait is not what makes DIVA special. The Department
of Defense has
been contracting with developers of those
technologies for years.
What's unique is the DIVA systems'
ability to communicate with each
other automatically and
intelligently in order to better detect and
then follow
individuals, according to Trivedi.
"Face-recognition systems
being developed by other groups are not
foolproof," said
Trivedi. "Sometimes they identify a person as X and
they are
not." Trivedi's DIVA architecture improves upon
identification
technologies by adding the elements of wide-area
surveillance
and the ability to adapt to dynamic events, so a person
is not
tracked solely on digital facial recognition.
To explain how
DIVA surveillance works, Mohan described what he
calls
"interesting events." An interesting event might, for instance,
be two massive objects colliding, stopping and then one of the
objects speeding away earlier than the system considers normal.
Upon observing this scenario, the DIVA system might report a
possible
hit-and-run to the police as it alerts other cameras
that it predicts
the runaway vehicle might pass. Those cameras
in turn anticipate the
car's path and continue notifying cameras
in various locations to
detect, identify and track the vehicle
until authorities stop it.
"The biggest challenge we have is
defining what is 'an interesting
event,'" said Trivedi, who
confirmed that the Department of Defense
would assist in
identifying visual cues and circumstances to trigger
the
intelligent cameras from casual observation to active
surveillance.
The watchful eyes of all those cameras
have raised the eyebrows of
privacy advocates.
"We see
something we warned about called 'mission creep,'" said Mihir
Kshirsagar, policy analyst for the Electronic Privacy
Information
Center.
"You have cameras that are supposed
to be monitoring traffic and red
lights," Kshirsagar said. "And
now they say, 'Let's look at the
people and crowds.'"
EPIC, along with other advocacy groups, is lobbying for
regulation of
unauthorized video surveillance, such as that
performed by DIVAs.
"Take the analogy of wiretapping … there
must be an audit trail,
procedures and rules for how you collect
and use that information,"
said Kshirsagar, who said such laws
do not exist for capturing video
images of individuals.
The only trail Trivedi is concerned with, however, is the
one his
cameras must learn to predict and follow. Since Friday's
research
grant announcement, he has added more researchers and
developers to
his team in order to build a powerful
indoor/outdoor camera system.
"Not just for lampposts, but
mobile," Trivedi said.
|
|
