Industry Professional
Transitioning from Video to Vision
IP-based digital capabilities have surpassed quality and edge processing capabilities
- By Vince Ricco
- Sep 01, 2018
In recent years, video surveillance has experienced a constant
and rapid evolution of both technology and use cases for IP
video in the traditional life safety and loss prevention worlds.
IP-based digital video capabilities have far surpassed predecessor
analog video systems in terms of both video quality
and edge processing capabilities such as recording, onboard
analytics and more. These advances have allowed the industry to
prioritize and focus on more useful video monitoring and recordings
based on items of interest, rather than blanket recordings
and time-consuming post-recording searches. Based on the added
intelligence, use cases well beyond traditional physical security
have been developing and are on the cusp of rapid adoption of
non-traditional uses for network video.
In the context of digital or networked video, the more common
terms we hear used are machine vision and computer vision. The distinction
between the two is that the camera is keyed in on monitoring
specific conditions, processes or items to look for anomalies, such
as a paint defects early in an automotive manufacturing process, or
ensuring the right pills are put in the right prescription bottle based
on visual verification in an automated prescription fulfilment process.
We are also hearing the term vision being associated with the terms
deep learning and artificial intelligence (AI).
Somewhere in the space between traditional video surveillance
applications and the promise of deep learning or AI, we are seeing
increasing deployments of smart systems, such as smart buildings,
smart parking and smart cities. The two examples detailed below illustrate
the potential power of video surveillance for providing deeper
vision when integrated into these types of applications.
Smart Parking
Smart parking is a prime example of the evolution of this vision applied
on top of traditional automation systems to increase communications,
efficiencies and profitability by augmenting available data.
A snapshot of smart parking without video would look something
like this:
Automated entry gate to provide access and a ticket to track parking
duration for billing. Potentially parking space-specific weight sensors
to determine which spaces are occupied. Potentially more generalized
pressure plates to count the number of vehicles entering or
exiting zones or parking levels to estimate the number of available
parking spaces. Exit gate with staffed and/or automated pay stations.
What we are now seeing is a vision overlay added to these nonvisual
systems, which provides added benefits. For example, video
overlay of parking spaces can allow for reservation of specific spaces,
as well as license plate recognition (LPR) of the vehicle for which
the space has been reserved. This can allow for tiered pricing based
on space location and can tie in to retail VIP or loyalty programs,
providing automatic notification to a retailer that a special customer
has arrived.
Additionally, LPR can be used at ingress points to identify entry
attempts by undesirable or blacklisted vehicles. Video can also be
used to identify vehicles that have parked in a manner that eliminates
the usability of an adjacent parking space.
It is important to note that edge intelligence like LPR capability
enables the recognition and transmission of key data points without
the need to transmit more exhaustive and very dense video traffic.
The increased efficiencies can be apparent and especially valuable
when tied in to a range of non-video sensors.
Largely depending on the environment for this use case, traditional
forensic video used to track theft, accidents and pedestrian
incidents are still attractive and valuable. As an industry, we are also
starting to see increased traction in adding additional video and audio
analytics to this architecture. These may include audio alerts in
an area of interest and automated alerts to security staff, police and
other first responders.
Smart Lighting
Another area where we are seeing video transition to vision in nontraditional
applications is the integration of video camera with smart
street lighting. Smart street lighting has seen a widespread increase in
integrated network communications based on technologies such as
cellular, Wi-Fi, wired connection and other low-bandwidth wireless
technologies such as ZigBee, Z-Wave, LORA and others.
The base need for these integrated technologies is to provide twoway
communication to control the lighting by turning it on or off
based on ambient levels, as well as to collect data from the lights,
including power usage and event trigger information—was the light
turned on based on motion, and if so, how often? You will now find
that many of the top street lighting manufacturers are not only embedding
the communications equipment but are also adding a measure
of IP video cameras. The promise of integrated video is the
ability to create additional services similar to those referenced in the
smart parking example, but also potential traffic analytics, such as
a vehicle in a bike lane, illegal parking, traffic flow monitoring, accident
detection and much more.
Smart parking and smart lighting are just two examples of use
cases where video is serving more as a vision sensor and data collection
technology. This is in addition to traditional video use cases
or exclusive of recorded video used solely to track event-based data
and report these data points back to a larger overall system. From
these brief glimpses of how video can transition to vision, potential
extrapolations from these use cases should immediately become clear.
For example, there is the promise of larger cloud-based learning
systems that can collect vast amounts of disparate data points,
including vision-based data in real time, and provide instantaneous
status and reporting on baseline anomalies. As a result, video surveillance
is moving from real-time reactive systems closer to more predictive
systems that can be deployed to solve operational challenges
while also helping organizations to mitigate, if not avoid, potential
security threats.
This article originally appeared in the September 2018 issue of Security Today.