How A.I. tracks crop growth in the U.S. — thanks to images from space

Here., a computer identifies wheat in Kansas from satellite imagery thanks to  the work of TEDxABQ speaker Steven Brumby (Photo: Steven Brumby)

Here, a computer identifies wheat in Kansas from satellite imagery thanks to the work of TEDxABQ speaker Steven Brumby (Photo: Steven Brumby)

Steven Brumby teaches computers to understand satellite data. The physicist aims to create “a living, breathing atlas for the world” using decades of satellite imagery analyzed through machine learning — in order to get a clearer picture of our planet, and the environmental threats facing it.

“We need to know the answers to some really hard questions,” Brumby says in a talk at TEDxABQ, “like, ‘Is that forest near my house going to burn down? Is the city in which I live going to run out of water? Is there a risk the regional food supply could collapse?’”

Maps can help, but they’re still underdeveloped, says Brumby. “Unfortunately, there are not enough humans in the world¬†that know how to do this map making … In the last six years, we’ve started to make maps of the food supply in the United States, but just for the United States, only once a year, and there’s nothing like this for the whole rest of the world … [for all] 7 billion humans.”

Brumbly has his program analyze raw satellite data like this to create maps (Photo: Steven Brumby)

Brumbly has his program analyze raw satellite data like this to create maps (Photo: Steven Brumby)

Fortunately, Brumbly’s come up with a way to speed up this map making — outsource it to the cloud. He created a program to teach computers to understand satellite imagery. “If you look at the raw satellite data coming back every day … it’s a mess,” Brumby says. “There are big holes because of clouds, and all sorts of noise. But when you show enough data to the computer, we can start to turn it into [maps].”

Brumby’s computers have created “what we believe to be the first true real-time cloud-free view of the world,” he says, learning to differentiate crops from bare soil, and even estimate how much food is coming out of U.S. fields. “For the first time, we’ve built a system that is faster, cheaper, and more accurate than the traditional manual techniques of estimating food crop across the United States.”

To learn more, watch his entire talk below:

Leave a Reply

Your email address and name are required fields marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>