Drones offer quick way to survey disaster zones

Drones offer quick way to survey disaster zones

A fleet of rapid-response drones could someday be used to assess damage immediately following a massive earthquake like the one that struck Mexico this month.

A University of Illinois graduate student headed to Mexico City today has developed a computer algorithm that can "see" and interpret structural problems in buildings and convey that information to emergency response teams.

The technology will also allow engineers to determine which buildings are safe to enter and which need to be inspected further.

Civil engineering graduate student Vedhus Hoskere, assisted by UI senior Michael Neal, will be taking drone video images of buildings in Mexico City to test the algorithm's ability to assess damage from the recent quake.

"His goal is rapid response during or after a disaster, to do a quick survey and assess where personnel and materials should be sent to most effectively respond," said his adviser, civil engineering Professor B.F. Spencer.

Hoskere will use the drones to do aerial reconnaissance, taking video images and using computer vision and machine-learning techniques to identify damaged structures, Spencer said.

Hoskere has already trained the drone's computer to recognize images of structural defects in buildings, such as corrosion, cracks or chips in concrete. The computer algorithm processes the images as digitized bits, just as humans process light and colors and shapes to recognize images.

"In computer vision, we train the computer to recognize that this is a bridge and there's no damage, and here's a concrete building and there are cracks all along the structure," said Spencer, who co-founded a startup company that developed wireless sensors to monitor strains on bridges and other structures.

Through machine learning, or artificial intelligence, the computer is then able to recognize new images that it hasn't seen before, from a real-life disaster.

The algorithm has performed well so far, and Hoskere hopes to get new video footage from Mexico City to assess its performance.

"We're trying to implement it in the real world, and see what the research gaps are that we have to solve," Hoskere said.

Testing a 'quadcopter'

The project blends computer vision and machine-learning technology with robotics, to make acquiring the images easier, Hoskere said.

"If we're able to acquire the data easily and process the data automatically, we'll be able to make informed decisions in the critical time after an earthquake much faster," he said.

Drones have been used for several years to take pictures of bridges and other structures for inspections, but they're flown manually, Hoskere said.

Hoskere is using a "quadcopter," a drone with four blades, that he will operate on site. But he's also developing ways to operate drones automatically, from afar, the first time anyone has combined that technology with machine-learning techniques, he said.

If he's successful, an operator could theoretically hit a launch button and send 10 drones on predetermined flight paths that could then stream back real-time video images. Upon their return, the computer algorithm could mine the high-quality video to identify where the damage is, Spencer said.

Spencer said disasters often affect wide geographic areas, and authorities need to know quickly where the hardest-hit areas are, which roads are open, and where to send paramedics and rescue crews. The swarm of drones could help them mobilize quickly and send people to the right places, he said.

'Engineer a way out'

It would also be faster and cheaper than using helicopters, as most cities don't have very many at their disposal, he said.

It's also difficult to get teams of structural engineers to a disaster site quickly to make those assessments, Hoskere said.

Though he grew up in India, he was born in Northridge, Calif., a year before the massive earthquake in 1994 that flattened highway overpasses and apartment buildings in the San Fernando Valley near Los Angeles.

His father is a structural engineer, and Hoskere has always been interested in both computer science and engineering.

The trip is being partially funded by the Department of Civil and Environmental Engineering, using gift money from the Will K. Brown Endowment.

Neal, president of the UI chapter of the Earthquake Engineering Research Institute, said he'll be providing logistical support.

"I think it's important for our department and our school to be present at these sorts of incidents and to be on the side of trying to engineer a way out of these sorts of problems," Neal said. "All these things we're learning about are things that will help us move towards a safer society."