Analyze This
Despite previous overhype, video analytics beginning to play bigger part in overall security industry
- By Brent Dirks
- Aug 01, 2009
Just like many organizations in these trying economic times,
the Miami-Dade County public school system is trying to
do more with less, even when it comes to security.
As one of the largest school systems in the country, district
law enforcement officials come into contact with many criminal
activities, including the usual trespassing and vandalism along
with gambling and fighting.
And one school in the district, Northwestern Senior High
School, is using technology that is beginning to percolate
throughout the security industry -- video analytics -- to help make
the campus safer.
ObjectVideo, along with partner Phoenix IVS, is providing a
solution that uses video analytics technology to convert video
data in real time through the detection, classification and tracking
of activities.
Powered by the company’s ObjectVideo OnBoard technology
that embeds video analytic technology within the cameras, the
system automatically alerts personnel to possible threats, eliminating
the need for someone to constantly watch video footage.
With the ability to monitor more areas with less manpower,
alerts are sent to officials both during and after school hours,
leading to the prevention of problems before they occur.
At the high school, the results have been tangible. The number
of incidents fell 34 percent in the 2007-2008 school year, and
officials are looking to install the technology throughout the district
over time.
“I’d like to have these in all our high schools,” said Charles
Hurley, interim chief of police for the Miami-Dade Schools Department.
“With the budget cuts and the fact that we just don’t
have enough police officers, this is the next best thing. It doesn’t
take the place of an officer, but it’s a heck of a nice complement.”
Realistic Expectation
According to Ed Troha, director of global marketing for ObjectVideo,
rollouts of the technology, like for the Florida school
district, should show that video analytics is working and making
its way in the industry.
“There are a lot of people who are saying video analytics
manufacturers are touting lowered expectations, and that is
wrong,” he said. “We think expectations of analytics should be
realistic, instead. We believe in taking a realistic look at what
can and cannot be done. And then we implement analytics as a
high-value ingredient to a solution that already exists or is
being planned.”
Being upfront with the customer leads to many advantages,
Troha said, including placing user expectations in line with what
they are willing to pay, along with lowering cost per channel.
“The practical, more straightforward approach is really working
out well for us,” Troha said.
Eric Eaton, CTO of Houston-based provider BRS Labs, said
simplifying the often complex functions of the technology will
help analytics spread to a wider audience.
“There is a lot of cautious optimism,” he said. “There are a lot
of people who want analytics to work. It’s obvious that there is a
lot of value added for a computer to go through tons of video
data. But the industry is in the process of maturing. Part of that
process is realizing what is real and what is junk.”
ioimiage’s president of the Americas, John Whiteman, echoed
Eaton’s sentiment about the industry, and he believes the technology
hasn’t reached its full potential
“I really don’t think the commercial marketplace and the
thousands of security dealers have really embraced the concept
and the technology,” he said. “The industry and our company
continues to grow at pretty quick and interesting rates. But it
hasn’t been exploited as some pundits had believed.”
But both manufacturers and dealers have work to do to help
push the technology, Whiteman said.
“We as an industry and manufacturers are being challenged
to educate everyone better,” Whiteman said. “On the dealer
side, there is an investment that needs to be made on their side
in training and certification and understanding the design.
Many dealers today would take a video analytic, plug it into a
camera and expect it to work. But the entire design needs to be
based around analytics.”
Different Approaches
Even though the video analytics industry is defined by a specific
task, companies in the space approach the issue with different
ideas to address the problem.
Troha said ObjectVideo is hardware and platform independent,
focusing on making analytics more easy and practical and
part of a system that a user may already have. To accomplish that,
the company embeds analytics libraries as part of a manufactured
product that the user is familiar with.
The company’s OVReady technology, an XML protocol, allows
any compliant video management platform to speak automatically
with any device that has ObjectVideo OnBoard enabled.
“That way, the analytics just becomes part of a system that a
user is already familiar with using and working with their trusted
manufacturer,” Troha said. “If the analytics just becomes another
part of that environment, we think this will make analytics
much easier to get at and use.”
Started in 2006, BRS Labs is a relative newcomer to the industry.
On the market since just September, the AISight software
program won the SIA’s New Product Showcase Video Analytics
award at ISC West in April.
Able to work with any PTZ camera, Eaton said, the product
takes a different approach from rules-based analytics.
“Our approach is not based on rules, but reason,” he said.
“Instead of trying to tell the computer what to look for, we added
a learning engine for the computer to learn by observation. It’s a
reasoning process with three phases -- observe, learn and respond.”
After a learning process, all behavior
observed by AISight is color-coded. The
most normal or typical activity is tabbed
purple, moving up a rainbow spectrum all
the way to red for the most rare behavior.
With the color-coding system, Eaton
said users can reduce time watching cameras
by focusing efforts on only observing
the most rare behavior observed by
the technology.
On the other end of the spectrum,
ioimage was one of the first companies in
the video analytics space. The company
sells IP video encoders and IP video cameras
with video analytics embedded on a
DSP in the device.
ioimage also was one of the first
companies to introduce a video analytic
edge device.
“Up until 2005, most people were promoting
software-based analytics running
on a server at a centralized location,”
Whiteman said. “Because of that, it was a
very sophisticated process to sell, design
and install video analytics. Also, because
of the cost, it was really confined to high-end
applications.”
But once the company introduced the
edge device, Whiteman said analytics
technology was able to increase its reach
across the market.
Looking At False Positives
While false positives are a common complaint
regarding video analytic technology,
the companies said that the problem is
getting better.
“When it comes down to analytics, the
two things people look for are almost-absolute detection, which
is almost impossible for anyone to provide, and zero false
alarms,” Whiteman said. “And there is a very delicate balance between
those two.”
Weather is a major factor in false alarms, Whiteman said.
“Whether its car headlights, shadows or bugs crawling on cameras
that replicate the size of a human at 250 feet away, the industry
has worked very hard to help negate those nuances,” he said.
Even though false positives will continue to be an issue, educating
the industry about proper expectations will help as well,
Troha said.
“It’s not working well for a lot of people because of expectations
coming in and that people are looking to solve problems
that can only be tackled on the TV show ‘24’,” he said.
With the learning capability of BRS Labs’ system, Eaton said
AISight looks at false positives in a different light.
“There is only a limited amount of data in the video stream,”
he said. “But if you looked at it, a false positive can be caused
when a rule is either too strict or too broad. With a system like
ours, the real question becomes what is the signal-to-noise ratio
of valuable data compared to junk you really don’t want to pay
attention to.”
Moving Past Hard Targets
Despite having roots in protecting hard targets and homeland security
functions, there is proliferation of video analytics being
used in other security and non-security applications.
“One of our biggest customers today is a national auto dealerships
with 80 to 90 locations across the countries with 100
channels deployed,” Whiteman said. “That is an excellent example
of how the technology is spreading across the industry. We’re
doing things in schools, hospitals and even high-end residences.”
Whiteman also said the technology is filtering down to such
mundane tasks as protecting a pool. With a camera protecting the
area, if a child gets too close, an alert is sent to a parent to check
out the situation.
Even past security, Troha said, analytics can easily be used
outside the realm of physical security.
“If you’re a retailer, the technology can do things like help
make sure that customers are not waiting too long in line and count
how many people enter a store buy something,” Troha said. “They
can not only save money but find money with video analytics.”
And even though the market for video analytics protecting
hard targets is maturing, Eaton said the technology can easily
spread across the security industry.
“I think there is broad applicability for video analytic technology,”
Eaton said. “A lot of the money and attention is being spent
on hard targets for obvious reasons. But technology that is used to
protect hard targets can easily be cross-pollinated for soft targets.
There are many installations out there where it is not going to hurt
for security guards to know the situation around them.”
At a New Level
By Kim Rahfaldt
Video analytics takes video systems to a new level by intelligently analyzing video images and
detecting whether certain predefined behaviors are taking place. However, how do you know
if video analytics is right for your application? Here are three examples where video analytics
plays a critical role in a security system.
Video Loss and Tamper
Video loss is the most important analytic a security system can capture. Perpetrators can be
creative in their efforts to avoid being caught on camera. They can place a cloth over a camera,
spray paint a lens, remove a camera from its location, cut the signal cable or remove the
lens altogether. And those are just a few examples. Whatever the method, once a camera is
tampered with, an alarm will be sent to notify security officers of the problem.
Obstruction of Scene
Another type of tamper event is the moving of a camera so it won’t capture illicit events. If the
scene changes, an alarm will notify security officers to review their monitors.
Motion Detection
Detecting motion is another key analytic function. Depending on the detection parameters defined,
an alarm will be generated if motion occurs within the camera’s view or other defined
window in the field of view. This is ideal for large fenced-in areas where no one is permitted.
Kim Rahfaldt is the public relations manager for AMAG. |
This article originally appeared in the August 2009 issue of Network-Centric Security.
About the Author
Brent Dirks is e-news editor for Security Products and Network-Centric Security magazine.