Watch for New Trends
Expect to see additional complex algorithms and more accurate detection methods
- By Brian Carle
- Dec 01, 2018
What technology and industry trends should you
keep an eye out for to stay ahead of the pack?
In short, complex analytics are growing hotter
which is made possible by the trend towards
GPU co-processing. Also, the country of origin
of security products is gaining more attention as tariffs and government
regulations draw attention to the geographic source of equipment.
More complex analytics algorithms, such as facial recognition, require
substantially more computational power than simple analytics, such
as cross line detection. Until recently this attribute has made more
advanced analytics and accurate detection methods impractical with
available hardware platforms. With the advent and popularization of
co-processing technologies, such as Graphics Processing Unit (GPU)
co-processing, processing large amounts of video in a short time has
recently become affordable and practical. As such, facial recognition is
starting to become a reality, and facial recognition products are now
penetrating the market even in commercial deployments.
Savvy consumers are incorporating advanced analytics such as
facial recognition to make better use of massive amounts of video
data. Facial recognition enables video surveillance to detect known
persons of interest attempting to enter a secure facility; a task that’s
nearly impossible for a person to accomplish. While a person can
certainly look for another individual, the task becomes impractical
when a long list of known persons of interest exists, and detecting all
of them is important.
Each consumer segment has its own common targets for detection.
Some examples of current and potential facial recognition applications
- Airports looking to detect terror suspects
- Casinos watching for known cheaters
- Schools identifying parents without visitation rights
- Retailers identifying known shoplifters
Generally, such deployments limit the number of cameras facial
recognition is running on, in order to reduce costs. Both the facial
recognition software and the corresponding hardware to process the
video can be expensive. Maximizing the ROI potential involves reducing
system costs by focusing the technology on areas where the
facial recognition is most likely to be successful. This can be done
by limiting the number of video channels being processed for facial
recognition to just the ingress/egress cameras, or cameras in key locations
(like security lanes at airports). Doing so minimizes the corresponding
hardware and licensing costs and maximizes the chances of
getting alerts on known persons of interest.
Other applications involve public and private partnerships. For
some time now, major sporting arenas hosting events have used facial
recognition software to compare faces of fans in attendance against
databases of wanted criminal suspects. Such a public/private partnership
could also be deployed in hospitals for the same purpose.
Search analytics are a category of analytics products which detect
and classify objects in recorded video. During an investigation, large
segments of video can be filtered more quickly by specified criteria,
harvested from the metadata in the processed video.
Such a product may allow for an investigator to import a sample
image of an object or person being looked for. Using the image sample,
the analytics will search the corresponding recorded video for a match.
Alternatively, a search analytic may detect and classify all objects
in a range of recorded video to allow filtering by category,
color and other attributes. For instance, an incident may have occurred
on a range of cameras over the past 36 hours. The last 36
hours of video recorded from the cameras located close to the incident are uploaded into the search analytic tool. Analytics are run to
identify and classify objects and record their properties. Later the
investigator can filter the video based on object type and properties.
If the investigator wanted to find all trucks, moving from left to
right in the scene, at a certain rate of speed and which are green in
color, the search analytics can now query its database of objects and
properties and call up video from the time matching objects were
detected. This makes the review process dramatically faster, but has
other benefits as well. During a lengthy review process, investigators
can become fatigued and lose focus, potentially missing key details.
Introducing search analytics reduces the likelihood of missing key
As with facial recognition, search analytics require substantial
computational power. The amount of computational power available
directly impacts the processing time, which can mean the difference
between getting results within an hour of the incident or a day later,
after video has processed. In situations where search analytics get
heavy use, or when time is of the essence, the hardware can add substantially
to the overall search analytics solution cost.
Also, with facial recognition, search analytics have benefited from
co-processing technologies such as GPU or FPGA based co-processing
integration. Such co-processing technologies allow a relatively inexpensive
co-processing card to be added to a server platform which
dramatically increases the speed of the analytics software running. It
is of note that co-processing technologies cannot be used with any
software package. Software needs to be written specifically to integrate
with a co-processing technology, so work with the software vendor
on the best hardware design.
Product Country of Origin
A recent trend with major influence on product selection, especially
security cameras, is the country of origin of the manufactured equipment.
This year, several factors have influenced the renewed focus on
equipment country of origin, including:
- Media within and outside the security industry has written on the
influx of cameras developed in nations not traditionally considered
to be allied with western nations.
- The United States has imposed tariffs on some imported equipment
from certain regions.
- The United States has signed into law the National Defense Authorization
Act for 2019, which specifically calls out by name particular
camera manufacturer’s equipment which cannot be sold to
U.S. government entities.
As such, buyers within and outside government organizations are
taking a second look at where their security equipment comes from.
Camera providers negatively impacted by these trends generally
provide lower cost products. Another impact as a result of this trend
which we may see will be reduced pricing pressure on other brands.
Whether one is in the planning stages of implementing a video security
system, has an existing installation or is in the process of an upgrade,
being mindful of near-future trends in the industry can help guide expenditures
and plan infrastructure builds. With the proliferation of
more accessible analytics-capable hardware, facial
recognition should increasingly factor in to video
security system builds. In addition, current political
realities may affect brand choices, and impact
camera hardware costs for future deployments.
This article originally appeared in the November/December 2018 issue of Security Today.