Trends at the intersection of video and AI
Installation
and use of CCTV Cameras for security & surveillance is a no-brainer.
Cameras are considered a fundamental commodity for setting up any surveillance
infrastructure, but at the same time, 24×7 monitoring of hundreds or thousands
of video feeds by operators doesn’t serve the purpose of providing proactive
surveillance and quick response to breaches.
2022, where most security cameras deliver HD performance, with more and more models offering 4K resolution with 8K on the horizon. Advanced processing techniques, with and without the use of infrared illuminators, also provide the ability to capture usable images in total darkness; and mobile devices such as drones, dash cams, body cams, and even cell phones have further expanded the boundaries for video surveillance. Additionally, new cameras feature on-board processing and memory to deliver heightened levels of intelligence at the edge.
But video has evolved beyond the capabilities of advanced imaging and performance to include another level: Artificial Intelligence. Video imaging technology combines with AI, delivers a wealth of new data, not just for traditional physical security applications, but for a much deeper analysis of past, present, and even future events across the enterprise.
This is
more than a big development for the physical security industry; it is a
monumental paradigm shift that is changing how security system models are
envisioned, designed, and deployed. Much of the heightened demand for advanced
video analytics is being driven by six prevalent industry trends:
1)
Purpose-built performance
Several
video analytics technologies have become somewhat commoditised “intelligent”
solutions over the past few years, including basic motion and object detection
that can be found embedded in even the most inexpensive video cameras. New,
more powerful, and intelligent video analytics solutions deliver much higher
levels of video understanding.
This is
accomplished using purpose-built deep learning, employing advanced algorithms
and training input capable of extracting the relevant data and information of
specific events of interest defined by the user. This capability powers the
automation of two important workflows: the real-time monitoring of hundreds or
thousands of live cameras, and the lightning-fast post-event search of recorded
video. Vintra video analytics, for example, accomplishes this with proprietary
analytics technology that defines multi-class algorithms for specific subject
detection, classification, tracking, and re-identification and correlation of
subjects and events captured in fixed or mobile video from live or recorded
sources.
2) Increased
security with personal privacy protections
The demand
for increased security and personal privacy are almost contradictory given the
need to accurately identify threatening and/or known individuals, whether due
to criminal activity or the need to locate missing persons. But there is still
societal pushback on the use of facial recognition technology to accomplish such
tasks, largely surrounding the gathering and storage of Personally Identifiable
Information (PII).
The good
news is that this can be effectively accomplished with great accuracy without
facial recognition, using advanced video analytics that analyse an individual’s
whole-body signature based on various visual characteristics rather than a
face. This innovative approach provides a fast and highly effective means of
locating and identifying individuals without impeding the personal privacy of
any individuals captured on live or recorded video.
3) Creation and
utilisation of computer vision
There are
a lot of terminologies used to describe AI-driven video analytics, including
machine learning (ML) and deep learning (DL). Machine learning employs
algorithms to transform data into mathematical models that a computer can
interpret and learn from, and then use to decide or predict. Add the deep
learning component, and you effectively expand the machine learning model using
artificial neural networks which teach a computer to learn by example.
The
combination of layering machine learning and deep learning produces what is now
defined as computer vision (CV). A subset but more evolved form of machine
learning, computer vision is where the work happens with advanced video
analytics. It trains computers to interpret and categorize events much the way
humans do to derive meaningful insights such as identifying individuals,
objects, and behaviours.
4) Increased
operational efficiencies
Surveillance
systems with a dozen or more cameras are manpower-intensive by nature,
requiring continuous live or recorded monitoring to detect and investigate
potentially harmful or dangerous situations. Intelligent video analytics, which
provides real-time detection, analysis, and notification of events to
proactively identify abnormalities and potential threats, transform traditional
surveillance systems from reactive to proactive sources of actionable
intelligence. In addition to helping better protect people, property, and
assets, advanced video analytics can increase productivity and proficiency
while reducing overhead.
With
AI-powered video analytics, security and surveillance are powered by 24/7
technology that doesn’t require sleep, taking breaks, or calling in sick. This
allows security operations to redeploy human capital where it is most needed
such as alarm response or crime deterrence. It also allows security
professionals to quickly and easily scale operations in new and growing
environments.
5) A return on
security investment
The advent
of advanced video analytics is slowly but surely also transforming physical
security systems from necessary operational expenses into potential sources of
revenue with tangible ROI, or as it is better known in the industry, ROSI –
Return on Security Investment. New video analytics provide vast amounts of data
for business intelligence across the enterprise. Advanced solutions can do this
with extreme cost-efficiency by leveraging an organization’s existing investment
in video surveillance systems technology.
This easy
migration path and a high degree of cost-efficiency are amplified by the
ability to selectively apply purpose-built video analytics at specific camera
locations for specific applications. Such enterprise-grade software solutions
make existing fixed or mobile video security cameras smarter, vastly improving
how organizations and governments can automatically detect, monitor, search for
and predict events of interest that may impact physical security, health
safety, and business operations. For example, slip-and-fall analysis can be
used to identify persons down or prevent future incidents, while building/area
occupancy data can be used to limit crowds or comply with occupancy and
distancing guidelines. In this way, the data gathered is a valuable asset that
can deliver cost and safety efficiencies that manual processes cannot.
6) Endless
applications
An added benefit is the ability to accommodate live video feeds from smartphones and common web browsers, further extending the application versatility of advanced video analytics.
Indian Army to be equipped with AI-based threat assessment software linked to national databases
