• 2 weeks ago
Dr. Robin Murphy, a researcher with Texas A&M University, was live on the AccuWeather Network on Nov. 1 to discuss new drone technology developed by those with Texas A&M that uses AI to map out storm damage quickly.
Transcript
00:00And after a hurricane tears through a community,
00:03a damage assessment needs to be done.
00:05It can be slow and costly, but a team at Texas A&M University
00:09has developed a way of using drones
00:11to quickly and accurately determine
00:14the amount of destruction to buildings and roads from above.
00:18Dr. Robin Murphy is leading the research team here.
00:22She joins us.
00:23Thanks for joining us this morning, Robin.
00:26Good morning.
00:28How are you using drones and AI
00:30to make faster damage assessments
00:32and how does your database work?
00:35Well, it works by a lot of fire rescue departments,
00:40emergency managers now have small drones
00:42that they can put up.
00:43So imagine particularly in a rural county,
00:46like we saw at Hurricane Helene or Debbie,
00:50where you've got a small community, 200, 400 houses,
00:54you can't get there, it's all flooded,
00:57pop up a drone and in less than 20, 30 minutes,
01:01you can get the entire area mapped out.
01:03But now what are you gonna do with it?
01:05You've got that, you can zoom in,
01:06but that takes usually days
01:08to figure out the amount of damage.
01:11So what we've done at Texas A&M is create an AI program
01:15that took 10 previous disasters
01:17and pull together damage and road assessments.
01:20And so does that autonomously in four minutes on a laptop.
01:25So now you're 30 minutes in the air,
01:27you drive up 30 minutes, you're down,
01:29and now you can put this in a laptop
01:31and get pretty accurate damage assessment.
01:34And that tells you how bad that community is,
01:37or maybe it's not hit as bad,
01:38maybe the predictive models are wrong.
01:41And so everything that we were thinking
01:42was gonna be bad is good, but if everything's really bad,
01:45now we've got to get more resources at there and there.
01:47And then you can also use that
01:49to get your damage assessments
01:50to the state and federal government
01:52to get the money to pay for all those recovery.
01:55Now, different colors show the level of damage.
01:58How does the artificial intelligence determine
02:01if there was no damage, partial damage,
02:04or complete destruction?
02:07Well, we gave them 10 disasters,
02:10drone imagery from 10 disasters,
02:1322,000 buildings, what, 400 acres of disasters,
02:20and trained it and showed each one put a polygon
02:23of where the buildings used to be and say,
02:26this is good, this is bad, gave those labels,
02:30and then it learned, and now you give it new images
02:33coming from like this, this is Florida State's team,
02:36they lead drone operations for the state of Florida,
02:40and get that data, and then it looks at it and says,
02:43oh, I've learned, I know what that looks like,
02:46and now I'm gonna give it one of these classifications.
02:49Now, you mentioned Florida,
02:51there's also states like Pennsylvania using this.
02:52Tell us how this is all being used and processed.
02:57Well, as Texas A&M, we work very closely
03:00with Florida State University,
03:01we're part of the Center for Robot Assisted Search and Rescue
03:04and we're also part of the Carnegie Mellon AI Institute
03:09on Societal Decision Making.
03:10So that lets us work with the Texas, Florida,
03:14and Pennsylvania Departments of Emergency Management.
03:16And so even though these techniques are still experimental,
03:20Florida has been quick to adopt it,
03:22and then Pennsylvania, during Hurricane Debbie,
03:25of course it became that tropical storm up there,
03:27and they saw never before flooding in rural areas,
03:30and they said, can we try this?
03:32And we're like, yes, yes, definitely so.
03:35And we actually tried it using a lower altitude,
03:38not drones, but crewed aviation,
03:41and it worked remarkably well.
03:43But it also showed there's a lot of difference
03:45between hurricanes in Florida and the Gulf Coast,
03:49and hurricanes when they go inland
03:51to rural mountainous areas.
03:53And so we've got a lot of research to do,
03:55and my great students, Tom Mancini and Priya Perali,
03:58have been working on that.
04:00And really quickly, have you done any damage assessments
04:04on Hurricane Helene or Milton from the last month?
04:08Well, we did Debbie and Helene with Florida,
04:12and Debbie with the tropical storm in Pennsylvania.
04:15Milton, it turns out they didn't use drones very much
04:17because that was a much more urban area,
04:21basically wall-to-wall crewed aviation.
04:23So you didn't need to pop up drones
04:26as much to go do these checks,
04:29which is why we really like these techniques
04:31because the rural areas, imagine you're understaffed,
04:35you've got a lot of area to cover,
04:37people aren't in big mega neighborhoods to check out,
04:40so now you can start using these small inexpensive drones
04:45and the software to figure out what's going on immediately
04:48rather than hoping the state gets satellite imagery
04:51or diverts something to fly over your county.
04:55So we're hoping it'll be a real game changer
04:57for the rural counties.
04:59Dr. Robin Murphy, researcher at Texas A&M University,
05:02thanks again for sharing that with us,
05:05a lot of great information.
05:06You have a good weekend.

Recommended