California wildfire camera network uses AI to spot new fires sooner

In short: California’s ALERTCalifornia camera network uses AI to flag possible new wildfires so dispatchers can check the video and respond sooner.

What's going on

ALERTCalifornia is a statewide network of cameras and sensors run from UC San Diego. The cameras watch remote, fire-prone areas, including parts of the Sierra Nevada. The network has grown to more than 1,200 cameras, although the exact number can vary depending on when it is counted.

The cameras stream live video to emergency responders. They can pan, tilt, and zoom, and they can do a full sweep of the area about every two minutes. At night, they can use near-infrared imaging (a way to see heat and low-light details, like night-vision).

Artificial intelligence, which here means software trained to look for patterns in images (like a tireless lookout), scans the camera feeds for signs of smoke or a new fire. When the system flags something, firefighters and dispatchers review the images, confirm the location, and keep watching conditions through the camera. UC San Diego says the AI tool, developed with CAL FIRE and DigitalPath, became available to all 21 CAL FIRE dispatch centers in September 2023.

ALERTCalifornia reports that in its first season the system detected more than 1,200 fires. It also says it identified fires before 911 reports more than 30% of the time. These are reported results, not a promise that every fire will be found, or found early.

What to watch

The key question is how consistently this approach works during the busiest parts of fire season, when smoke, haze, and weather can make video harder to interpret. It also matters how well the AI alerts fit into dispatchers’ daily workflow, since humans still have to verify each possible fire.

Source: NYTimes