Fire, Camera, Action
From the rafters of an energy plant, a new type of camera is on guard. Two men enter its field of vision. One lights a fire. Within seconds, a red-square appears around the fire’s image on the monitor connected to the camera. From the command center, security personnel receive an alert signal and visually verify on their monitors that a fire has broken out. They then contact the fire department.
The camera has, in a sense, replicated the way a human eye communicates with the brain. It not only saw the fire, it recognized it as fire, and alerted security of its presence. But how?
“The underlying technology is software algorithms,” according to Mac Mottley, CEO of AxonX, makers of the SigniFire IP Camera.
To explain, Mottley asks a question: “How do you know what a fire looks like when you look at one?” Before I can answer, he says, “You identify it because it’s a really bright object, it flickers according to a certain pattern, it smokes, it expands at a certain rate. We almost model the software algorithm on how humans try to look at it.”
Casey Grant, program director at the Fire Protection Research Foundation, which sponsors research into innovative antifire technologies, says AxonX’s camera is an example of video imaging or video analytics.
Video-imaging systems harness the power of algorithms to construct neural networks, similar to the neurological structure of the human brain, which allows the camera’s software to make a decision based on what it observes.
When placed in a fire-detection context, says Grant, video-imaging technology offers certain advantages. For instance, large building spaces have long frustrated fire detection experts. Such spaces have usually relied on either sprinkler or suppression systems to guard against fire. But oftentimes, the fire has to reach a particular size to activate the systems, making property damage from flames, heat, and smoke more likely. Furthermore, sprinklers can damage whatever is stored in the space, while suppression systems can leave that space uninhabitable for humans after the suppressant has been released.
Video fire detection (VFD) systems, such as AxonX’s SigniFire IP camera, can help a facility avoid both direct damage from the fire and associated damage from extinguishing systems by ensuring early detection. Because the system relies on visual analytics, it can identify a smoke plume before it has time to waft up and set off a conventional spot-type smoke detector. With its ability to recognize flames, it can also identify a fire before it gets large enough to kick up enough smoke to set off smoke detectors. This process allows security personnel to respond to the fire before it gets out of control and put it out without having to resort to a facility’s sprinkler or suppression system or the fire department.
The system also cuts down on false alarms—or at least makes it easier to quickly assess whether an alarm is valid. When a spot-detector or sprinkler system goes off, there’s no way to verify that a fire set the instruments off without going to that location. “Usually what happens is that someone gets sent out to check to make sure it’s a fire before they call the fire department,” says Mottley.
VFD systems, like all detectors, do sometimes generate false alarms. Distinguishing between an actual fire and other nonfire events such as conventional building lighting, hot work, or lightning can be challenging, says Grant. But at least, with the VFD, “we start streaming video, and people can identify it as a fire very quickly,” he says.
UTC Fire & Security, another company developing a VFD system, has created a fusion algorithm designed to cut down on false alarms. Essentially, it analyzes data from many different individual algorithms, and the system won’t sound the alarm unless all the criteria for fire and smoke are satisfied.
“For example,” says Alan Finn, a research fellow at UTC, “if the system sees certain combinations of motion and color that mimic the characteristics attributed to fire, such as an orange leaf moving in the wind, but other important factors—like smoke—are not present, the system won’t alarm.”
Another advantage of VFD systems is that they also offer surveillance capabilities, giving companies the ability to watch over a location for unwanted intruders as well. Understanding the convenience of this two-in-one solution, AxonX’s camera allows users to set up motion-detection zones that will alert the command center when someone moves through a secure area. “You get speed of response, and you get situational awareness,” says Mottley.
And by combining fire protection with video security systems, companies could save money. “When you start combining security with fire detection, then you’re killing two birds with one stone,” says NIST’s Kerber.
To gain trust in the marketplace, new technologies, such as VFD, must satisfy the relevant National Fire Protection Association (NFPA) fire codes. The most important code for VFD systems to satisfy is the National Fire Alarm code, or NFPA 72. The code not only governs how the product performs but also how it is installed, tested, and maintained, says Lee Richardson, staff liaison for NFPA 72.
According to NFPA 72, VFD systems must be “listed” or “approved” by a product-testing company, which means the product meets certain standards. The two most recognized in the United States are Factory Mutual Global (FM Global) and Underwriters Laboratories (UL). The former is more concerned with testing fire and smoke detectors in commercial and industrial settings, while the latter tests smoke detectors in a life-safety setting, such as hotel rooms. In January, FM Global listed AxonX’s SigniFire system as a fire-detection system suitable for use in commercial and industrial applications.
Because the system uses a video camera, the company doesn’t see the “technology in hotel rooms, bedrooms, and the like, which are dominant in the life-safety market,” he says. “We see a lot of interest by commercial and industrial applications that have large volume, high risk, or high asset-protection requirements. Therefore FM is more applicable and desirable to us than a UL listing.”
Customers in a life-safety setting interested in using SigniFire must first satisfy their relevant fire codes. The system can then be added on for more redundant detection of flames, smoke, and intrusion, according to Lynch.