Ease of Inspiration
From the practical solutions today to tomorrow’s app store
- By Robert Muehlbauer
- Oct 01, 2012
Inspiration can come from many avenues. A personal
experience. Frustration with the status
quo. A friend. A chance encounter. Sometimes
even from our dreams. The trick is turning that
inspiration into action.
The world of video analytics has suffered from
an inspiration-to-action dilemma since September of
2001. To be fair, it was the unrealistically high expectations
that were set for analytics that made action
nearly impossible. There were dreams and promises of
Minority Report-like video intelligence that could predict
a crime before it happened. Fortunately, industry
expectations have come back down to earth, and we’ve
learned some valuable lessons along the way.
to say the video analytics industry has been a
failure this past decade would not only be wrong, it
would be insulting. Security, operations and loss prevention
professionals are using a cavalcade of analytics
to great success, ranging from basic to advanced to
what would seem futuristic to most.
Basic Analytics
Video motion detection. Nearly every IP-based digital
surveillance system today uses some form of video
motion detection. Whether VMD is enabled to protect
specific critical assets or better control bandwidth and
storage, motion detection proves its worth. Traditionally,
video motion detection has relied on pixel-based
change in an image, but the latest algorithms actually
act as true motion detectors, improving accuracy.
Tampering alarms. Likely the most underused
and underrated analytic of the bunch, a tampering
alarm alerts security guards if the camera has been
manipulated in some way. The camera learns the
scene and can tell if the pixels have been changed—
as when someone covers or spray paints the lens,
or even turns the camera to a different position.
The camera can also alert the system if it has been
knocked out of focus or if it’s not sending a video
signal at all—but unfortunately not many users have
set up this simple configuration.
Audio detection. While laws and regulations across
many regions of the United States prohibit the recording
of audio, the use of sound detection is widespread.
This analytic detects if a sound reaches a certain
decibel level and alerts the system owner. In its
elementary form, it’s perfect for catching late-night
prowlers. And while audio detection is classified as
basic here, developers have added features to move
this analytic more into the advanced category with
what could be classified as Audio Detection 2.0.
Some municipalities are using audio coupled with
PTZ cameras to detect gunfire and initiate the camera
to automatically shift its field of view to where it
heard the sound. Even more advanced analytics are
hitting the market to detect threat levels in a voice by
analyzing the audio pattern to determine a person’s
stress levels.
Advanced Analytics
Cross-line detection. Drawing virtual trip wires drawn
in the camera’s field-of-view seems like a simple task,
but the trick is to configure and identify an object’s direction.
In low-traffic environments, this application
runs inside the camera or encoder to detect when an
object crosses a virtual line while moving in a specific
direction set by the user. It’s perfect for building entrances,
loading docks, parking lots, roads that abut
school zones and restricted areas.
Heat mapping. Unrealistic expectations were a big
reason for analytics’ false start, but the need for near-perfect
accuracy was another. If a bag left behind is missed,
there could be disastrous consequences. In the world of
store operations and marketing, however, there’s more
room for error—and more room for growth.
Retailers, while lagging behind other verticals
when it comes to IP video adoption, are enjoying
novel uses of analytics. Specifically, heat mapping can
track customer movement around the store and create
an overlay to show hot and cold zones with traffic
patterns. This data is invaluable for merchandisers
and can give the marketing department reason to take
interest in the surveillance system.
People/object counting. People counting is another
analytic prevalent in the retail world, and applications
such as Aimetis’ People Counter are becoming
increasingly accurate as more advanced algorithms
are installed on more powerful cameras. For example,
retailers can use it to determine staffing levels during
peak hours. Universities are using the same technology
to count cars entering the parking garages on
game days to better manage capacity, and, like retailers,
some K-12 schools and nonprofits use people
counting to prove that more funding is needed to staff
afterhours or special events.
Future-is-Now Analytics
License plate recognition. LPR-specific cameras have
been around for years, but they are expensive and
analog-based. Today, LPR analytics—like the one recently
released by ipConfigure—are designed to run
inside a high-resolution network camera, making the
camera usable for multiple surveillance needs while
opening the doors to additional applications. Typical
LPRs will still be used in toll collections, parking
garages, campus environments and city surveillance
applications, but emerging applications can be found
in gated communities, after-hours check points, longhaul
container tracking, visitor management, gas station
reward programs, and even automated will-call at
restaurant curbside carry-outs.
Fire and smoke detection. Yes, smoke and fire detectors
have done their jobs for years. But in large,
indoor expanses, it can be difficult to pinpoint the
source of the fire as smoke drifts to several surrounding
detectors. Analytics that leverage video verification
of smoke and fire, such as Fike’s SigniFire, can
alert operations to the source so it can be quickly extinguished
while the building is evacuated.
Video synopsis. This is an application you need to
see to believe (and to fully understand). The leader in
the video synopsis space is BriefCam, whose software
detects events throughout a predetermined period of
time—say, 24 hours—and then overlays all the events
on top of one another to enable security professionals
to browse hours of video incidents in a matter of
minutes. To help police operations and surveillance in
airports and cities keep track of all that’s going on
in the scene and pinpoint a specific event, time signatures
appear over the object in motion to indicate the
time of day.
Behavioral analytics. Behavioral analytics like the
ones installed by BRS Labs in San Francisco and
Louisiana’s Greater Lafourche Port Commission
might seem like a Minority Report-style prediction
analytic, but, just like a person observing a scene, the
analytic learns the normal and routine activity from
the day-to-day video data it captures to identify when
something doesn’t look right. It still requires the use
of an operator to determine if what the camera detects
as out-of-the-ordinary is in fact suspicious, but it
calls the operator’s attention to actionable video and
keeps eyes fresh and minds sharp.
The Cutting-Edge
In only a few years, the saying “There’s
an app for that” has become cliché. If
you haven’t said the phrase, you’ve likely
rolled your eyes at someone who has.
But the saying exists for good reason.
We have seen much inspiration from
developers of software applications using
the Apple iOS and the Android platforms.
When we face a challenge in our
everyday lives, we run to the Apple App
Store or Google Play to find a solution.
Why can’t there exist that ease of inspiration
in the surveillance world? Unfortunately,
the list of real-world surveillance
analytics pales in comparison
to the 40 billion-plus apps downloaded
onto smartphones and tablets.
Yet the concept of the network
camera or encoder mimicking an app
store model is not new. Embedded
motion detection has been running
inside the camera/encoder (i.e., “at
the edge”) for almost 10 years. More
applications have become embedded,
including tampering alarms, cross-line
detection, audio detection and local
storage capabilities.
Not until the emergence of the latest
application-specific integrated chips
(ASIC)—that is, a processing chip specifically
designed for the surveillance
world—has the potential been this
great. Following the path of Moore’s
Law, the processing power of these
ASICs has been doubling every 18
months, and we now have enough processing
performance to open up a new
world of application opportunities.
What are the advantages of running
an application at the edge? First,
traditional analytics can require many
CPU cycles, which limits the number of
cameras supported per server. Offloading
some of those CPU cycles onto the
camera or encoder itself and processing
all or some of the video at the edge reduces
the cost of the total solution.
Second, the solution will be far more
scalable as less space is required in the
server room. Cameras and encoders can
be updated to meet the latest firmware
and features and moved to different hot
spots, depending on specific data needs.
Third, the surveillance customer
may not have the option of having a local
server where the camera is installed,
but devices can be remotely accessed
when needed without any additional
hardware required.
As IP cameras continue to become
increasingly powerful, the guiding vision
is to create a surveillance app store
that inspires developers to look at this
market as a target for their talents.
Video analytics was once the great
promise of digital video to make our
systems more intelligent and our personnel
more efficient. Until that is achieved,
we need to think outside the traditional
surveillance world, share our unique
challenges with each other and turn inspiration
into action. If not, when the
time comes that you see a revolutionary
yet surprisingly simple analytic solution,
you’ll be asking, “Why didn’t I think of
that?”
This article originally appeared in the October 2012 issue of Security Today.