Where does Artificial Intelligence fit in the security industry?


Despite much discussion and hype, most people are unclear as to what constitutes artificial intelligence (AI). This is particularly true in residential and commercial security, where it has picked up steam the past few years. Furthermore, how related terms like analytics, machine learning and deep learning are subsets of, cross over or diverge from AI is a mystery to the majority of security pros and customers alike. At the same time, those folks would likely assume that robotics performing sophisticated tasks would necessitate some degree of AI.

“AI is a broad term used to describe anything that appears to ‘think’ like a human brain,” says IntelliVision’s Tony Lacy-Thompson. “Machine learning is a technique whereby a computer can improve its ability to do a particular task by processing large amounts of data about that task. Deep learning is a subset of machine learning that uses neural networks. A single layer of a neural network will process the data once and pass its result to the next layer, which does its task and passes its answer to the next layer. This is known as a convolutional neural network or CNN. Video surveillance analytics is an excellent application of AI and deep learning, enabled by the advent of GPUs [graphics processing units].”

Certainly video analytics is one of the most prominent ways AI has been applied to the security mission, and by extension physical security information management (PSIM) systems that integrate a multitude of security data sources to create situational awareness. One of the most sophisticated video analytics comes from BriefCam.

“Our Video Synopsis and deep learning solutions make surveillance video searchable, actionable and quantifiable,” says company CTO Tom Edlund. “BriefCam’s video analytics platform is built on a unique fusion of computer vision and AI technologies empowering new and innovative safety, security and operational efficiencies. It recognizes and extracts objects, along with information about the type and attributes of those objects. We continue to further improve accuracy and gain new and more granular classifications and attributes.”

Another noteworthy innovation that has been brought to analyzing video is IC Realtime’s Ella. The Cloud-based deep learning solution augments surveillance cameras with natural language search capabilities. “From an industry standpoint, we’re just now diving into the possibilities of AI and deep or machine learning,” says IC Realtime CEO Matt Sailor. “With Ella, and the advancements being made in machine learning, users can search an entire library of recorded footage using the phrasing and keywords that comes naturally to them. This solution has been designed to become more intelligent over time and better at determining what video clips are most important to the user.”

Imagine durable and agile robots being equipped with such types of advanced algorithms, processing and intelligent sensors. It opens up a vast new security landscape in which human threat is minimized and remote locations and perspectives heretofore inaccessible become available for real-time analysis, evaluation, and response.

Eventually, as the devices become more refined and prices plummet, the applications will be vast. In the meantime, varieties of AI are likely to infiltrate most aspects of our daily lives.
“Machines taking over low-cognitive tasks will be the big trend for years,” says Milestone Systems’ Keven Marier. “Amazon is applying this to retail stores where the checkout concept is being replaced by customers just walking out. This type of thinking and tool creation is in its earliest infancy but will continue to address problems and add value. The intelligent industrial revolution is happening all around us. It will be very disruptive, but also insightful and liberating.”

It’s not quite yet all systems go for this market. There remain technology, cost and other obstacles. For example, advances in AI and robotics are outpacing the government’s ability to implement sensible standards and legislation. “The biggest limitation of unmanned ground and aerial vehicle technology is regulation,” says Genetec’s Sean Lawlor. “Looking at UAV technology as a security solution can be complex. For example, if we built a system allowing an automated flight plan of a UAV with a camera mounted on it to traverse a security perimeter, we would need to know if during its operations it would ever violate a local law. Understanding when and where regulations influence UAV and UGV usage will become critical as their security use increases.”

Perhaps the biggest hurdle AI-powered robotics faces is fear of “Westworld”-like glitches, particularly when dealing with something as critical as security. As with furthering any solution, managing expectations is imperative. “False positives/negatives are a substantial problem for most computer vision solutions,” says Umbo CV CEO Shawn Guan. “Making sure the device is responding to what is actually happening is crucial and a challenge. AI is a technology, not a magic wand. Vagueness around the term coupled with bombastic media reports makes this challenging. If someone is expecting the stars, then even delivering the moon is a disappointment.”