In the early days of thermal technology, military and law enforcement were virtually the sole markets for thermal cameras. It would have been astonishing at that time to fast-forward and see the various jobs that thermal cameras are doing today. Thermal cameras are on school playgrounds, in hospitals, at zoos, on highways, within oil refineries and many more places, proving to security integrators that they are a utilitarian and cost-effective solution to consider for their customers’ needs.

But, not surprisingly, many of the applications for thermal aren’t security-related at all. As new technology is adapted to thermal cameras, it has created many new uses that spill out beyond security. Artificial intelligence (AI) coupled into thermals, in particular, has led to a pivotal time and great growth for these cameras, says Steve Burdet, manager, solutions management, Axis Communications, Chelmsford, Mass.

The idea of traditional thermal technology in surveillance is to see in challenging lighting or complete darkness. Is there a car at the perimeter of the property? Is there a person climbing the fence? “But you would need a human watching it to see that it’s a person,” explains Aaron Saks director of product training, Hanwha Vision America, Teaneck, N.J. “We’ve had analytics, but that analytic couldn’t tell the difference if it was a deer or a person. It just sees a hot object on the scene,” he says.

“A traditional, pixel-based analytic could never tell you, ‘this is a human,’” Burdet adds. “It would do its best guess, but to get there it’s so, so hard. This just allows you to really classify objects in a much better way. AI … absolutely improves the ability to detect those more complicated things that we just couldn’t before.”

AI Gives Greater Purpose to Thermal Cameras

Because thermal cameras can only be viewed in grayscale, they obviously can’t use an analytic that detects certain colors. However, AI confers many other benefits to thermals. In addition to giving a major boost to detecting objects — yielding more precise detection with fewer false alarms — it also expands the detection distance.

“Within the camera, detection analytics have advanced from simple motion detection to a fusion analytic that couples motion detection, neural network classification databases, and behavior analytics into a single analytic that not only performs at a higher level of accuracy, but also increases the detection distance capabilities,” explains Matt Strautman, director of business development, FLIR Security, a division of Teledyne FLIR, Hudson, N.H.

“When you start talking about … the introduction of advanced analytics or AI, that’s when you start talking about some of the newer ways that you can detect people or use AI for some of the industrial applications,” says Michael Poe, director of product management for video and business intelligence, 3xLogic, Fishers, Ind. “You can put a little bit more of a predictive analysis on, say, a piece of equipment to know that on Tuesdays, it gets a little bit hotter because the demand is a little bit more, so don’t notify me then.”

AI-fueled analytics furnish organizations with a great amount of data collected by the thermal camera. “What does AI allow? In a lot of ways, it allows the data that creates a ton of new opportunity,” Burdet says.

Simple data points generated by an AI-enabled thermal camera help it distinguish between a human and a non-human object. Depending on the application, that could be the most relevant information needed.

But end users can go much further. Analytical data can be used to plot the trail that an object took while it was in the camera’s field of view, for example. This “positioning” metadata then can be worked in different systems, such as in industrial applications.

“If our cameras … are just the detector of an object, the analytic is saying now that you’ve detected an object, I want to analyze what that object is and then I’ll determine if it’s something I care about or not care about,” Burdet describes. “Being able to generate this as an object and have the software on it or any other things basically begin their action is a really big step forward with AI, [as is] having that data just be used for data purposes [without the need for] any recordings.” He clarifies that the latter scenario isn’t a life-safety application, but purely for process monitoring or something similar.

Further, collected data can be analyzed within a business intelligence platform. If it shows that the thermal camera helped prevent a business problem, then end users can urge the organization’s leaders to make future investments in the technology, Burdet says. “Creating that data is super, super helpful. You didn’t really get that with traditional analytics,” Burdet says.

Temperature-Detecting Thermals Put to Use

There are two main types of thermal cameras — those tuned for object detection and those tuned for measuring exact temperatures, called thermographic or radiometric. The latter can be used simply to produce a notification when a temperature exceeds a specific threshold that would indicate a fire is present. “More advanced methods correlate the thermal image to an analytic that then analyzes the rate the temperature rises to better determine if a fire is present, Strautman explains.

“Looking beyond traditional security applications, thermal cameras are being used to provide solutions to improve life safety measures, specifically around early fire detection. Thermal cameras now offer a bi-spectral solution that includes a high-end, uncooled thermal camera along with a 4K visible camera. The radiometric versions of these cameras can produce temperature data on the objects they target, and when coupling these two streams with specific behavioral analytics we create smoke and flame detection analytics,” Strautman describes of this type of solution.

There are many things that produce heat in the industrial world, including electrical equipment like that found at substations, power-generating facilities, hazardous sites such as garbage or recycling areas, and other industrial processes. Operators know the expected temperatures of their equipment and if a thermal camera detects it to be hotter than it should be, then they can immediately make adjustments to prevent fire or explosion.

Thermal imaging does well at industrial sites — where parts prone to overheating, welded pipe bends, loose joints, malfunctioning valves or blocked pipes can not only disrupt operations, but possibly create a life-safety hazard. Burdet points to garbage sites, for example, where heat builds up in the piles. If operators can be alerted to spots where the temperature is rising quickly, they can request that the pile be sprayed or knocked down to prevent a fire.

“I believe with AI and the mature market that we’re in, people are looking at thermal as a potential solution to some of their process-monitoring problems that before it really couldn’t because your algorithms would have to be too complex,” Burdet describes. “So now it’s actually achievable in a way that people are finding good ROI on it.”

Burdet says that the complexity of the algorithm for such applications made it prohibitive until recently, because either the file size was too large to “live on the edge” and required a connection to the cloud or because it wasn’t accurate enough. This is relevant, he says, because edge processing with AI means the camera only has to send the data that’s necessary.

“If I have to run it to the server or even to the cloud — which is just more bandwidth and a further distance to travel — I have to send the full video stream even if nothing’s happening,” Burdet says. “And that means that the cloud or the local [server] has to constantly be analyzing nonstop in the hopes that something might be occurring. That’s just a lot of computation; so to get the best of all of those worlds, you want to leverage edge devices. It has no transmission, no lag, no latency, no whatever. If it says something occurred and as long as it’s accurate and you can rely on it, you can begin your reaction instantly.”

From Easy to Complex Uses for Thermals

In addition to the wide category of process monitoring, there are many other applications for both types of thermal cameras. FLIR’s Matt Strautman says that security integrators who have embraced thermal cameras are continue finding opportunities where thermals support both life safety and operational efficiency.

When thermal uses both AI and perception technology it then has wide use in emergency response and personal safety, describes Art Stout, director of product management – Artificial Intelligence Solutions, at Teledyne FLIR, Goleta, Calif. Systems using these technologies can detect people entering restricted areas or find missing people with a drone.

“AI is also used to add video processing of hydrocarbon gas emissions, enabling oil and gas extractors and processors to quickly detect harmful leaks to protect the environment,” Stout says.

“Thermal imaging-based AI object detection is helping to significantly improve autonomous vehicle capabilities, including automatic emergency braking (AEB), to more effectively detect people and large animals at night and in other low-visibility scenarios where traditional visible cameras struggle,” Stout adds.

In urban environments, thermal cameras often are used to monitor traffic conditions, especially on dark roads and in tunnels.

Michael Poe at 3xLogic thinks of applications for thermal technology as belonging in three buckets: workplace safety, which aligns with the process monitoring applications described by Burdet and Strautman, with a focus on employees to prevent accidents or injuries; healthcare, which puts thermal cameras in hospital rooms to remotely monitor patients’ temperatures or at entrances to monitor visitors; and life-safety, which could be locating people in a fire. “A lot of firefighters have thermal viewers on their helmets to help them see in smoky environments,” Poe says.

Separate from those three disciplines, thermal cameras are ideal for what some people term “environmental intelligence.” Wildlife conservationists are using them to monitor large swaths of land by helicopter or drone for the presence of animal groups, Poe says. People tasked with researching climate change are using thermals for forest fire detection.

Some thermal applications are “kind of cool stuff,” says Aaron Saks at Hanwha. He describes an installation in which a thermal camera is observing a dam from a distance. Dam officials know the temperature of the concrete on the structure, so if the dam were to breach or if water was pouring over the dam, the temperature would change and generate an alarm, Saks says.

Another is a zoo, where certain animals are sensitive to infrared lighting, so traditional surveillance cameras with built-in infrared illumination can’t be used. Thermal cameras allow the animals to be observed at night without impacting their rest cycle, he says.

Years ago, mines would buy thermal cameras, Saks explains. This was because miners would carry various battery packs for powering equipment. They found that when the batteries failed, their temperatures spiked before exploding, which is a critical event when working underground. Therefore, thermal cameras would monitor the health of those battery packs, Saks says. Today, with the rise in electric vehicles and their battery-charging stations, being able to monitor those temperatures is also a job for thermal cameras.

Burdet cites a school district where the floor material is made of shredded rubber tires, rather than the more traditional wood chips. If someone were to toss a cigarette butt on the rubber material, it could — and did — start a fire, he says.

“They were using thermals because they had so many parks burned down that they needed to be able to detect more quickly when these things caught on fire, which is a very specific problem to this town, but nonetheless it shows that there are a lot of applications when people are willing to talk to their customers about what they’re truly trying to solve,” Burdet says.

“If you understand what thermals can provide for you, you might realize that they could be a perfect fit as opposed to trying to jerry-rig something with visual cameras or some other detecting technology that might not be as useful,” he says. One way to think about it is that thermal cameras are optimal at detecting at much further distances than a traditional camera, as long as they have a clear line of sight, he says.

More to Cheer About

AI and advanced analytics aren’t the only developments that have put thermal cameras in focus for solving problems. Other improvements that have made a difference are the availability of bi-spectral cameras that feature both a thermal and a visible camera in one unit.

“With a bi-spectral camera having both a visible and a thermal sensor in the same camera, being able to use the thermal sensor to detect a person and the visible high-resolution camera to put the image on an operator’s screen opens up a lot more applications, Saks says.

Thermal cameras traditionally have been a premium product with a relatively high price tag. Not so anymore, as both their size and cost have decreased in recent years while at the same time their resolution has increased. “Thermal cameras used to be very large and very expensive; [now] they’re getting smaller, lower cost, and that’s allowing us to deploy into more places,” Saks says.

Most thermal cameras are subject to strict export control regulations due to the concern of cameras getting into the wrong hands or being used for nefarious purposes. However, “Thermal cameras that have a frame rate of under 9 frames per second do not require end user export documentation,” Saks says. Many companies now sell thermal cameras meeting this spec, which means it can avoid the regulations that otherwise require completing special documentation and meeting challenging logistical requirements.

“What we’re seeing is companies are deploying these, I’ll say lower-end, thermals — but now you … don’t have to worry about a huge amount of paperwork and all this effort just to sell one. So that opens up a lot of applications,” Saks says.

What was once a niche product is now more accessible and useful to the market, as evidenced by the wider number of thermal camera models available from manufacturers. Technology developments such as AI analytics, smaller size, better resolution, and ease of purchasing have made this possible.

“A couple of years ago most people still viewed thermal technology as a higher end solution for perimeter protection. Now thermal solutions are being viewed as necessary for perimeter protection, and more and more integrators and end users are inquiring how they can use thermal for life safety and operational applications,” Strautman summarizes.

“Traditionally thermal cameras were to detect people coming across a perimeter, whereas now we’re using them to detect things,” Saks says. “And, because prices are coming down and they’re easier to procure now, they’re being used in a wider range of applications. A lot of it is customers are coming to us with ideas that we never thought of or weren’t cost-effective to do whereas now, ‘Okay, sure, sounds good.’”