https://x.com/i/article/2112532628968554496

【X 长文】How can researchers use AI to spot wildfires?

Did you know we can now detect early-stage wildfires from space before they spread? FireSat, a new satellite constellation from @EarthFireAll, in partnership with @GoogleResearch and @MuonSpace, uses AI to spot wildfires earlier and more accurately than ever before.

How do the satellites work with AI to fight fires? Let’s break it down:

To confidently spot fires, you either need massive, expensive satellites, or a network of smaller, low-cost satellites that use AI to make up the precision difference. So we got to work building custom cameras optimized for fire detection and trained our AI models by flying them over controlled burns, giving the system a reliable baseline of what a true fire looks like from above. By tracking how these scenes change over time, the FireSat constellation can spot fires as small as 5x5 meters — that’s basically the size of a car!

Standard satellite images are tricky to decipher. A reflective cloud, a smokestack, or even a backyard grill can look like a wildfire from so far away. Plus, existing imagery is often up to 11 hours old, which doesn’t help when you're trying to stop a rapidly spreading fire.

For first responders, time is crucial, and detecting fires while they're still small allows crews to contain them, better forecast how they might spread, and keep communities safe. More reliable and frequently updated satellite imagery is a crucial step in wildfire identification and containment.

The first batch of operational satellites is already in space, and more are being added over the next few years. By 2030, it will be possible to have roughly 50 satellites in orbit, taking a complete picture of the Earth’s surface every 20 minutes.

Watch the full video to learn more: goo.gle/4i7lJ6m