Changing the Look of Security
Analytics can make searching video possible in real time, and what this means for you
- By Jennifer Hackenburg
- Oct 01, 2019
Video analytics have been around almost as
long as CCTV itself, rescuing security operators
from the nearly impossible task of continuously
and effectively monitoring video surveillance
feeds. The concept of video analytics
is simple enough: algorithms allow a device to analyze video in
real time and send alerts to the user.
Analytics can also categorize aspects of recorded video so it is
more easily searchable. While modern technology has made analytics
more accurate and more affordable than ever before, we are
still seeing some barriers to adoption. Here is how video analytics
are changing the way we look at security, and how you can leverage
this to provide more value to your customers.
History of Video Content Analysis
One of the earliest developed iterations of video analytics was
motion detection. While video surveillance systems did correctly
identify suspicious motion within a scene, such as that of an intruder,
they were also great at identifying tree branches blowing
in the wind, or rain falling in front of the camera lens. The prevalence
of false alarms gave video analytics a bad rap, and rightfully
so. Like the boy who cried wolf, video analytics were so unreliable
that camera operators started ignoring them–defeating the purpose
of having analytics at all.
Advances in hardware and software greatly improved accuracy,
but for a time, the highest-performing analytics were only available
in the most expensive equipment. Fast forward to today, where analytics
have become mature enough that we are seeing them creep
into the consumer surveillance market (see Nest, Ring).
On the camera and recorder side, better hardware lets devices
react to event detection faster and more accurately. An increase
in processing power allows for more sophisticated analytics at
the edge. Inexpensive recorders work in conjunction and provide
sophisticated suites of analytical capability. Now the only thing
keeping end users from adopting video analytics is education.
Breaking the Barriers to Adoption
One of the most basic barriers that prevents users from incorporating
analytics into their security system is lack of understanding.
“Video content analysis” gets lost amongst buzzwords like
“Artificial Intelligence (AI)” and “deep learning.” It is important
to distinguish video analytics from the more advanced forms of
AI. Technically, AI means “getting a machine to mimic human
behavior in some way.”
Most video security equipment today uses rule-based analytics,
and at the edge (embedded into cameras). This form of AI
does not improve their analysis and accuracy through continued
examination of data.
Rule-based analytics are a cost-effective solution that enables
the system to examine many events and objects simultaneously in
a way that would be impractical or impossible for humans to do.
However, rule-based analytics do not have the cognitive ability to
interpret activity as well as humans can.
The prevalence of video content analysis in procedural forensics
crime dramas on TV has also negatively impacted the practical
application of video analytics in real-life situations. The hopes an avid crime show watcher may have for their video surveillance
system can be unrealistic. This “CSI effect” means that integrators
should set expectations early on and focus on educating the
user what the system can do instead of cannot do.
How Analytics Improve
Return on Investment
Given the innovative abilities of today’s surveillance systems, users
should be looking beyond simply getting a clear picture. They
should be striving for a system that arms them with intelligence
during the process of capturing video. This means not only installing
devices with analytical abilities but learning how to use
those functions as well.
A simple sales pitch is that video analytics improve the ability
of a surveillance system to detect specific events in real time, and
permit the user to easily search footage upon playback. Utilizing
video analytics has an array of benefits: making the system operate
more efficiently, reducing manpower requirements, providing
business intelligence (such as improving customer service or
better understanding staffing requirements), hastening forensic
analysis and enhancing detection accuracy.
All of these attributes add up to increased ROI for the user,
who can get more out of their security system and share the security
budget across multiple areas of the business. This is an
incredibly significant selling point to someone who is investing
thousands or even millions of dollars into video surveillance.
Analytics at the Edge
Thanks to the hardware improvements and more accurate algorithms
mentioned earlier, many highly-effective video analytics
are available at even the lowest price points. Among the analytics
that are most widely available are camera-based analytics, also
called edge analytics. These are particularly well-suited to challenging
locations or budgets because they require lower bandwidth
and fewer servers. In addition, they are easily implemented
in real time. The line crossing / tripwire function is one of the
most commonly deployed edge analytics.
Expanding upon line crossing is perimeter protection, where
analytics support detection of invasion in a defined area. Wellsuited
for industrial parks, school campuses, factories and warehouses;
perimeter protection employs algorithms that differentiate
humans and vehicles. For example, enabling an object filter
and selecting “human” will tell the system to send an alert when
humans are detected and ignore vehicles and animals. Another
filter could detect if vehicles are entering pedestrian areas and
vice versa.
Today’s perimeter protection solutions have achieved better
accuracy than previous versions, where frequent false alarms
were caused by environmental conditions or wild/stray animals.
People counting is another easy-to-use edge analytic. The
operational capabilities of people counting give users valuable
data to run businesses, such as discerning how many people are
gathering in a given area and even keeping track of how much
time they have been standing there. By setting rules to trigger an
alarm if, say, three or more people stand in line at the grocery
store for more than three minutes, store managers can be instantly
alerted when they need to send more staff to the cash registers
and reduce wait times. Using people counting at the entrance to a
nightclub or bar lets security staff know when the establishment
is approaching the capacity defined by fire code.
Search and Review
Metadata – a set of data that describes and gives information
about other data – collected through video analytics helps categorize
and organize footage so that it is easier to search. When
there is no known event to search for, surveillance systems produce
a seemingly endless amount of camera footage and historical
data that is not viewed at all. Now, video analytics allow for
certain attributes to get tagged to the footage before it gets stored
on the recorder. For example, face capture functionality can also
store facial attribute metadata – such as gender, age, expression
and whether the person is wearing glasses and/or has facial hair
– along with a snapshot of the face. Once such metadata is introduced,
smart search is possible.
Smart search opens a world of possibilities in locating very
specific footage. Operators can find, say, vehicles detected in a
tripwire area on a certain evening, or use the facial attributes
function to display all clean-shaven people with glasses who
passed through a particular train station in the past week. This
will yield a thumbnail gallery you can examine in minutes rather
than hours. Shortcuts like these let users efficiently review the
data they have and glean deeper information from it.
Video analytics have given rise to an exciting advancement
in playback: a time-compressed view where moving objects are
tagged with the time of day, such as BriefCam’s VIDEO SYNOPSIS
and RapidRecap from Lorex. A report of a thirty-minute
sequence, for example, generates a video synopsis that is only 53
seconds long because the view simultaneously displays events that
have occurred at different times. People and vehicles that move in
and out of the field of view are superimposed on a stationary
background along with a time stamp “affixed” to them. It’s a fascinating
glimpse into the future conveniences video analytics may
have yet to offer.
It is Up to Us
Video analytics are truly changing the way we use security. They
turn our security cameras into data collectors that help businesses
run more smoothly. They save time by only giving us the information
we need, whether we need to know about it as soon as it
occurs, or after the moment has passed.
Now that analytics are less likely to produce false alarms and
are more widely available in budget-friendly products, a growing
number of end users can benefit from them. It
is up to us, the manufacturers and installers,
to clear up misconceptions and show the true
value that analytics can provide.
This article originally appeared in the October 2019 issue of Security Today.