Real World Opportunities
Historically, there were two things you could do with surveillance video—watch it or record it to watch later. Video analytics changed all that. Video content can now be analyzed using various computer algorithms to identify specific user-defined content and to trigger an alarm or response automatically. The range of video analytics functionality can include identification, behavioral analysis and situational awareness.
- By Bill Taylor
- Dec 01, 2011
With smart functions incorporated into today’s network video
cameras, images can be analyzed before they leave the camera,
which can then trigger an alarm or relay other data to enable a
system-wide response. Alternatively, high-quality, higher-resolution
images supplied by a camera can travel along a networked
system and then be subjected to even more sophisticated video
analytics at the server level.
In either case, the resulting system provides functional advantages
that far exceed the view-or-record scenario that historically
limited the effectiveness of video systems to the attention span of
the operator.
In the last decade or so, while some in the surveillance industry
were only beginning to embrace smart video, the technologies
surrounding video analytics have continued to evolve and mature.
What was once often dismissed as a technology that was difficult
to install and plagued by false alarms has now become a useful,
mainstream tool in the IP video system. Making the transformation
possible are better cameras, more computational power—
both in the camera and at the server level—and greater ease of
use. Video analytics today are more robust, more dependable and
more easily incorporated into the IP network environment.
Smarter Cameras = Better Images
The old computer acronym GIGO—garbage in, garbage out—
also applies to video analytics. The quality of the image being
analyzed has a critical impact on the effectiveness of the system.
Fortunately, the images now being supplied by new, smart network
cameras are more detailed and better than ever. Megapixel
images are providing greater resolution, which translates into
more-detailed images for analysis.
Enhanced image processing also helps overcome lighting
challenges to produce usable images in a variety of real-world
situations. Cameras can now provide a clear view of an entrance
where complex lighting combines dark and light areas, and highresolution
images can render an automotive license number perfectly
legible.
One challenge is outdoor scenes, which are especially difficult
due to lighting and seasonal changes. As edge devices in a video
analytics solution, cameras can optimize the video image for better
picture detail. For example, image processing can expand the
dynamic range of an image to overcome lighting challenges that
undermine the effectiveness of video analytics. In some applications,
it can be difficult to identify the details of dark images.
However, with video captured by smarter high-definition cameras,
it is easier to identify the face of a person, to see what is being
stolen, or to read a license plate.
The bottom line is, a better image equals better analytics. Improved
video quality also extends to Face Wide Dynamic Range,
specifically designed to enhance the details of a face for better
identification and analysis.
In-camera video motion detection (VMD) functionality is
now more sophisticated than ever, offering multiple programmable
detection areas, multi-level sensitivity adjustments and multistep
detection sizes. Smart cameras also offer a “privacy zone”
function to mask private areas, such as house windows and entrances/
exits. Intelligent resolution can prevent deterioration of
an image during digital zooming; and variable resolution technology
makes it possible for a less-important part of an image—such
as the sky—to be coded at a lower resolution to save data file size.
Features such as wide dynamic range, adaptive black stretch,
and 3D digital noise reduction make images clearer and more detailed;
progressive scan provides images with no motion blur or
tearing; and H.264 full-frame-rate video can be recorded at the
camera level using an SD/SDH memory card.
All of these expanded features enhance the opportunities to
use video analytics. Integrators must be attentive to the angle of
the camera, camera positioning and lighting. Camera setup can
be simplified by use of auto back-focus technology to provide
optimal image quality and focus with a single push of a button.
On-board Analytics
Embedding technologies into edge devices—as opposed to using
centralized platforms—makes it easier for integrators to install
and set-up video analytics solutions. Intelligent network cameras
now offer face recognition, advanced motion detection and auto
tracking, among other functions.
Video analytics inside cameras on the edge of the network can
identify objects left behind or track customer traffic patterns or
count crowds. These functions at the camera level can be integrated
into systems that provide additional functionality.
More on-board intelligence empowers additional and more
effective video analytics. However, the centralized approach has
its own advantages. Chief among these is a collective database,
along with system-wide real-time control and management of
functionality and alarm response. A combination of centralized
intelligence and video analytics at the edge works well. Intelligence
inside the camera helps to minimize the system’s computational
load and the amount of data that travels across the network,
which makes for better use of network infrastructure, while
core capabilities are still centrally managed.
System Considerations
Some integrators may have been intimidated by the complexity of
programming video analytics systems, which often provide many
configuration choices. Integrators have tended to avoid the extra
expense and time involved in completing these more challenging
installations.
Early installations of video analytics systems were beset by
system crashes and other problems—and the time lost in figuring
out what went wrong. Integrators had to be precise in installing
these systems because one error could undermine the whole system.
Also, installation typically involved many separate components
from various suppliers. Integrators had to be mindful of
factors such as computer requirements, software updates and
security patches.
Currently, the move toward smarter edge devices has made
it easier to install video analytics. These systems require less
configuration because edge devices—cameras—have on-board
intelligence. Because analytics are already installed and pre-programmed
into the device, there are fewer trouble-shooting steps,
and the systems are much easier to configure. In short, intelligent
video cameras provide an entry-level opportunity for integrators
to get comfortable with analytics without having to deal with
higher system complexity.
More intelligence in the edge device can help pre-select and
filter what video is shared across the network. Using the network
infrastructure to view only selected video can help to minimize
bandwidth and storage requirements, which can be a concern
with systems that incorporate megapixel cameras.
Newer network cameras also can detect faces automatically,
even in high-contrast lighting situations, and even if there
are multiple people in a frame. The use of Face Wide Dynamic
Range functionality ensures a clear image of a face, and a facedetection
function detects the face’s position. In some configurations,
the camera detects the face and sends metadata to an NVR,
where the metadata is analyzed and compared to an existing database.
Upon a positive face match, alarm notification can be
sent, and the image can be displayed. An NVR with a real-time
face-matching function works in tandem with a smart camera,
comparing any faces detected by the camera against a database
of previously registered faces.
Higher resolution works with in-camera intelligence to ensure
more detailed metadata is available for face-matching capabilities.
In-camera analytics also extend to identifying the gender of
a captured image and other variables.
A Range of Applications
From retail to transportation, education to homeland security,
there are many diverse applications for smart camera technologies.
In the retail environment, smart cameras can be used to
detect a habitual shoplifter among several customers entering a
store at once. Another application might be to identify and alert
management when a VIP customer enters a business. Identifying
customer traffic patterns, sounding an alert when a check-out line
is too long or ensuring that customer service practices conform
to a standard are all applications that can benefit from smarter
video cameras. Beyond the security and loss prevention benefits,
the technology provides quantitative improvements to business
operations than can impact the bottom line.
In homeland security or large stadium applications, smart
cameras as edge devices can enable video analytics applications
such as identifying objects removed or objects left behind. At
schools and colleges, motion-triggered cameras can identify campus activity after-hours. In critical infrastructure environments,
video analytics could specify virtual trip wires and provide an
alarm when someone crosses the designated border. In transportation/
gas station applications, high-resolution images make it
possible to capture a license plate number, whether viewing archived
video or using license plate recognition (LPR) software.
In public areas and town center applications, clear color images
are available 24-hours a day through the fusion of high sensitivity
and adaptive digital noise reduction with auto back-focus.
In a banking or financial application, cameras can provide a
clear view of an entrance where the lighting situation is complex
due to a mixture of dark and light areas. A megapixel image,
combined with image processing to maximize the details despite
variable lighting, enables precise identification and efficient bank
operations. Face recognition can provide an alert when a known
criminal enters the bank.
In short, the possible applications of video analytics are just
now being realized. Focusing on a broader array of benefits beyond
security is an opportunity for suppliers to help customers
enhance cost justification strategies. Video and data can be integrated
with other applications such as retail systems, human
resources, process management and access control systems.
Putting “Smarter” to Work
Objects all around us are becoming smarter every day as computer
intelligence shows up almost everywhere in our daily lives.
Smarter machines are made possible by microchips and highdensity
microstructure processing technology, and we see daily
examples of how embedded intelligence can transform the functionality
of everyday things.
In the case of video surveillance, the ability to use computer
algorithms to analyze and interpret the content of video presents
a new level of system functionality, provided the capability is
strategically employed in the broader system environment. Once
considered systems in themselves, video analytics can today best
be thought of as a capability that can be incorporated into an
IP-based networked system. Video analytics operating as part of
a larger system provide a host of benefits to boost functionality
and to overcome the limitations of a system’s human component.
Video analytics can alert operators to actual incidents, eliminating
or reducing the need for individuals to watch hours of video
just waiting for something to happen. Using analytics, a system is
more effective and provides greater security.
However, video analytics is a tool, and as such does not apply
to every situation. Thoughtful and selective application of this
powerful technology will likely be the largest factor to ensure its
continued and growing success in the future.
For the video surveillance industry, embedded intelligence
opens a host of possibilities centering on system functionalities
and interoperability. When you factor in advancements in highdefinition
imaging, the future of networked video surveillance is
even brighter. Better images and a heightened capability to analyze
the content of those images provide a powerful combination
that can transform how video systems are used. The two technologies
work together in ways that are greater than the sum of
the parts, and surveillance system integrators and users reap the
benefits of the resulting improvements in system functionality
and performance.
This article originally appeared in the December 2011 issue of Security Today.