Technology increasingly makes it possible to identify wildfires early and before they get out of control.
Networks of video cameras watch wildlands for the first hint of fire, and now artificial intelligence (AI) is being deployed to interpret data from those cameras and alert personnel about an imminent wildfire sooner than ever.
FireScout AI
FireScout is an AI program seeking to address the issue of wildfires. Using AI, the program can create awareness and “understand” smoke from 10 to 25 miles away, based on the video from a digital camera set up in the field.
In a test conducted with utility Pacific Gas and Electric (PG&E) in California, the technology was able to detect 200 out of 300 fires, each in less than a minute.
Golden moments
In the case of FireScout working with a network of video cameras, that period has been reduced
“Golden moments” are the time that lapses between when a fire starts and when it is reported, a critical time when a potentially catastrophic fire can be averted.
In the case of FireScout working with a network of video cameras, that period has been reduced. In some documented cases, notification preceded the 911 call by minutes or even a couple of hours. The program also can detect fires at night when everyone is asleep. In remote areas where cameras are located, there may be no one around to call 911 anyway.
 |
| As for the Coyote Fire near San Diego in 2022, FireScout AI provided notification more than 2 hours in advance of a 911 call. |
AI-enabled smoke detection
A fire-spotting camera sends a digital image to the AI cloud, where advanced AI systems detect smoke and then send an alert for confirmation. The "human-in-the-loop" monitors the system to observe false alarms and to confirm a positive alert. Human-confirmed fires populate a FireScout dashboard and trigger an alert by SMS and email.
A 3D map view allows users to have better topographical information, and users can share the response status with their colleagues and stakeholders.
early detection of wildfires
FireScout has worked with public utility companies, individual businesses, and high-net-worth properties to increase early detection of wildfires.
The technology is based on AI analytics software developed by Alchera, a 10-year-old public company in Korea. The company operating outside Korea is called Alchera X and includes the FireScout AI program to detect wildfires.
Video cameras for wildfire detection
In addition, there are several hundred cameras installed outside the U.S. in Korea, Australia, and (soon) Canada
Currently, there are about 1,000 video cameras installed to watch out for wildfires in California, with more cameras to be added. There are also 300 to 400 cameras watching for wildfires in other states, including Washington, Oregon, and New York. In addition, there are several hundred cameras installed outside the U.S. in Korea, Australia, and (soon) Canada.
Beyond California, FireScout is working to minimize wildfires in Australia, and beginning now in Canada, the early steps in an expected global rollout.
Over a billion training images
In February 2023, the head of the Korean Ministry of Interior and Safety visited the United States to discuss installing Alchera X FireScout AI software to resolve the wildfire issue plaguing their country.
FireScout has the most accumulated user data among AI systems with over 10 million exclusive and independent images of smoke, clouds, and dust storms. They have more than a billion total training images accumulated for AI education. The system detects smoke or fire as soon as it is visible in the camera frame.
99.9% accuracy
FireScout’s “human-in-the-loop” process ensures 99.9% accuracy. Given the high rate of false positives, humans are deployed to verify a fire before sending first responders.
If a camera detects smoke suggesting a nascent wildfire, the FireScout system sends a short snippet of video to allow a certified human monitor to see where and how much smoke there is. Monitors are carefully trained and have familiarity with the geography and various types of smoke.
Visual information
The video snippet provides our human experts with visual information they can analyze"
“When we can save property and lives, that’s our objective,” says Michael Plaksin, Alchera X’s Vice President of Sales & Marketing.
“We can detect smoke when it is almost invisible to the human eye. The video snippet provides our human experts with visual information they can analyze.”
fire suppressant chemical
FireScout works with Mighty Fire Breaker, a company that provides a fire suppressant chemical that can be applied to brush or to property and/or houses.
High insurance costs are driving companies out of California, and such a fire-suppressant strategy can lower insurance costs.
Helicopters for fire response
In addition, the use of helicopters, alerted by text message or email can respond to fires. Helicopters located throughout the state of California can drop water on a fire within about 10 minutes.
Helicopters include “personal size” smaller helicopters, operated by remote control, and even Blackhawk military-size helicopters. FireScout works with RAIN (Rapid Aviation Integrated Needs) Fire to deploy services faster than other firefighting organizations in response to fire reports.
Challenges with AI
AI struggles to differentiate between actual smoke and a business smoke stack or even snowmaking equipment
“When you are looking at smoke, we don’t like having a margin of error,” says Plaksin. Over time, as AI becomes smarter, it might one day be possible to approach 100% accuracy using only AI.
Currently, AI struggles to differentiate between actual smoke and a business smoke stack or even snowmaking equipment in a ski area, for example. Including humans in making the final determination is a critical element.
Successful fire mitigation
“The good news is that we are having an impact,” says Plaksin. “I can’t tell you how excited I am about that. We are helping to mitigate the fires, although you can’t prevent them. If we can minimize the acres that burn, it can be a huge help.”
“We relieve in the accuracy of our detection, and we have been extremely successful detecting and protecting,” says Plaksin.
