The sheer amount of video and other data generated by security and non-security devices today is massive, and its growth is seemingly exponential — with more and more created every day. Given the challenge of collecting, let alone analyzing it all, it’s no surprise that much of what video has to offer goes by the wayside because there’s simply no easy way to deal with it.
But that thought is quickly becoming outdated, thanks to the rise of artificial intelligence (AI), particularly its use in video surveillance, says Marc Tanguay, product marketing lead, Arcules, Irvine, Calif.
“All of this data has tremendous potential for analysis and integration with other systems to offer valuable insight into both security and business operations,” Tanguay says. “AI has emerged as a technology that can streamline this process of constantly learning and adapting to real-world scenarios, as well as enable better business decision-making.”
For instance, he says, AI can learn what a normal environment should look like based on continuous video monitoring of a designated area. In the case of anomalies to the normal environment, AI technology would classify those anomalies or irregularities as outside of that norm.
Narrow AI using machine learning and deep learning is the technology deployed today, as opposed to general AI, says Paul Garms, director of regional marketing, North America for Bosch Security and Safety Systems, Fairport, N.Y. Machine learning is a way of training an algorithm by feeding it data as a method of teaching it to adjust itself to improve its performance. Deep learning is a more complex form of machine learning based on artificial neuron networks (ANNs) that mimic the structure of the human brain.
“The applications of machine learning and deep learning in video security vary widely,” Garms says. “You can now teach a camera to recognize situations, objects and even clustered objects by providing sample images or defining parameters.”
The most common applications today for AI-powered cameras are the same analytics that have traditionally been employed, such as loitering, intruder detection or entering/exiting an area, says Aaron Saks, product and technical manager, Hanwha Techwin America, Teaneck, N.J.
“AI becomes a powerhouse when used to eliminate false positives from shadows, foliage or animals, by only triggering the analytics when the correct type of object is detected, such as a person or vehicle,” he says. “A deeply integrated AI solution allows for additional metadata search parameters to speed forensic investigations, by entering search criteria such as clothing color or if they have a bag or hat.”
The primary draw of AI in video is its ability to turn video feeds into streams of actionable data, says Eric Hess, senior director of project management, SAFR, Seattle. This makes the technology an affordable way to add enhanced situational awareness.
“With no extra manpower you can effectively monitor more video feeds, extract more information from those feeds, and be more confident that you haven’t missed anything important,” he says. “The central goal of security teams is to protect people and spaces by preventing costly losses and dangerous incidents from occurring. If you have the right information in real time and can stop an emerging incident from occurring then you’ve succeeded.”
Applying the Technology
Video object detection using AI is a highly accurate method for analyzing activity around a home or business, delivering detailed information and sophisticated actions.
“Our AI-driven system detects people outside and can activate a series of light automations to mimic someone being awake or present even when they’re not,” says Anne Ferguson, vice president of marketing, Alarm.com, Tysons, Va. “This creates a more effective deterrent for intruders and informs the homeowner of activities of interest, while also reducing unnecessary alerts caused by other moving objects like animals or foliage.”
Combining AI, video and intrusion control creates an opportunity to heighten awareness of who may be trying to arm or disarm a system at any given time.
“One other valuable use of AI is facial recognition when disarming a panel,” Ferguson adds. Customers can receive an unexpected activity alert whenever the face of the person disarming the panel does not match the face of the person assigned the access code that was entered.
Naturally, the most common applications for AI in video are related to security. For example, software is trained to detect when a perimeter is breached but AI-powered algorithms can help determine whether it’s an actual threat — like an intruder — versus an animal or other object. Beyond this is the ability for AI-enabled software to help improve business intelligence to streamline processes. There are a number of use cases in the manufacturing world where video was used to determine best practices and target areas where time could be improved.
“In these instances, time is money, and utilizing this technology can help pinpoint areas in the assembly line where holdups occur and allow managers to brainstorm a better way forward,” Tanguay says. Similarly, he adds, using AI-enabled video to determine usage within a building is a trending use case.
“Traffic in a certain area of a building, such as a retail store showrooms, can be analyzed to outline potential choke points that are having a detrimental effect on the flow of people and assets,” Tanguay says.
Garms adds some more creative and interesting uses of AI include detecting the build-up of ice on the openings to tunnels, detecting water levels above a certain threshold, and detecting state changes. Other examples include detecting a vehicle attempting to travel across a railway crossing while the gate is down, and detecting oil levels in storage tanks.
“We are finding new and interesting applications for AI nearly every day,” Garms says.
How Prevalent is AI in Video Today?
According to Ferguson, AI is helping drive a higher video adoption rate and as a result, its inclusion with video continues to climb. With AI available through both its residential and commercial service packages, new Alarm.com accounts often include the deep learning video platform from the start while an increasing number of existing customers are adding that feature to their accounts.
“If compatible cameras are already in place, service providers can add analytics remotely, reducing the need for truck rolls while boosting their RMR and significantly improving the customer experience,” Ferguson says. “Our expectation is that AI will be an increasingly important feature of video monitoring in the years to come.”
For uses within physical security, AI is still relatively new and catching up with the capabilities already available in other markets, but that doesn’t mean it’s not growing at a high rate, Tanguay says.
“So many manufacturers tout the ability of their products and solutions to incorporate AI-driven capabilities, when in reality, motion detection is what is being used today,” he says. “In all honesty, in video surveillance, we’re only scratching the surface of what’s possible in the day-to-day analysis and collection of information into actionable insights.”
According to Garms, real AI capabilities have only been available for about a year now.
“The adoption of AI is occurring much faster than that of its predecessor video analytics. Since video analytics have become more robust and trusted, integrators are now more eager to take the next step and move into AI,” Garms says.
As cameras are becoming more powerful, AI applications are increasingly becoming edge-based.
“Until recently, AI applications were much more suited to centralized compute models which required massive compute resources to gain these analytics and insights. We believe we will see more and more AI bundled solutions at the edge moving forward,” says Sean Lawlor, data science team leader, Genetec Inc., Montreal.
There are plenty of edge- and cloud-based AI systems on the market today, but it is often only in specialty camera models or too expensive to be used across the board, Saks says, adding that that may be about to change.
“What we are seeing in 2020 is the adoption of a new generation of devices with purpose-built SoCs (System on Chips) to handle AI processing. This allows broader AI deployment so that entire video systems will be able to take advantage of the benefits of AI versis only having a few key cameras enabled,” he says.
The technology may have grown by leaps and bounds in recent years, but there is still a large opportunity that hasn’t been fully tapped, Hess says. Advances in computing power and edge-based intelligence are making these security enhancements more and more affordable — and thereby more accessible — to a wider variety of end users.
“We’re on the cusp of this technology being available on plug-and-play cameras so a mom and pop shop with one or two security cameras or a small, resource-constrained security team can leverage computer vision to make their environments safer with a low TCO,” Hess says. “The future possibilities are really endless. When the billions of dumb cameras currently deployed around the world are transformed into intelligent devices, we can enhance security and convenience virtually anywhere. We’re far from the tipping point in terms of this technology being deployed for all the security and business benefits it can offer.”
Advice for Working With AI
AI helps drive more meaningful video interactions and allows property owners to extend perimeter security, stay connected to family and catch the unexpected. These are the main points integrators can make with their customers to earn their buy-in on this largely misunderstood technology.
AI also offers tremendous opportunity for integrators to expand their offerings beyond physical security into the realm of improving business operations and more for their customers, Tanguay says. Cloud-based AI systems can simultaneously deliver video security and business intelligence, and integrators can deliver greater value by educating their customers about the benefits of applying video data beyond physical security applications and transforming video into business intelligence.
When it comes to evaluating AI solutions, Lawlor says the most important thing is to ask questions.
“We see many providers promising various levels of performance. However, these are often based on results that were measured in a lab environment, with ideal lighting, etc.,” he says. “It’s important to ask the vendor to explain how their solution will work, and what levels of accuracy can be expected in your specific environment and use case. If they can’t give you a precise answer or can’t refer to someone in the organization who can, you may want to look elsewhere.”
And considering how AI has been portrayed in movies and TV, one of the first things integrators must do is set realistic expectations for customers, even as the technology continues to advance in the future.