This new technology that detects deadly landmines with 92% accuracy could help Ukraine
New technology to detect lethal explosives designed to maim and kill has been tested by researchers from the Demining Research Community, a non-profit organization that bridges academic research and humanitarian demining efforts.
The researchers have been in Oklahoma for two weeks setting up grids of mines and munitions to train a drone-based, machine-learning-powered detection system to find and identify harmful explosives without the need for people to do so, Scientific American magazine reported on Wednesday.
"It's not easy to find places where you can lay dormant mines and fly drones. We're trying to automate the detection of different types of landmines, anti-personnel mines, anti-tank mines, and we're doing this through remote sensing and drones," the researchers said.
"Currently, most humanitarian demining is manual, so this is a beep and prod method where we have metal detectors, layout grids, and you're able to scan a metal detector really low down to the ground and poke it to see if there is a mine there."
The research team is led by co-founders of the Demining Research Community Jasper Baur and Gabriel Steinberg at an explosive-ordnance disposal field laboratory operated by Oklahoma State University.
"Convolutional neural networks are scarily accurate and a lot of applications that are better than humans. We've achieved a 92 percent accuracy with the limited amount of data we've gotten in this field season" said Steinberg.
Mines and unexploded cluster munitions are designed to prevent enemy troops and vehicles from using roads and fields. The issue is that unexploded cluster munitions and mines do not "turn off" when a war ends. Instead, they pose a lethal threat to civilians for decades, sometimes outlasting the countries that deployed them.
There are millions of active mines and munitions scattered across dozens of countries. Baur and his colleagues hope to make their drone-detection system available to demining organizations worldwide to help with efforts to make post-conflict countries safe.
Can this technology help in Ukraine?
On Thursday, the United States announced $89 million to assist Ukraine in clearing landmines.
Roughly 13 percent, 160,000 square kilometers of the Ukrainian territory of land may be "contaminated" by landmines and other unexploded ordnance, an area roughly the size of Virginia, Maryland, and Connecticut combined, according to Ukrainian government officials.
"The neural network detects landmines just from the data you give it. So we give it varied data and lots and lots of labeled data. That's why we're out here to collect lots of real-world data." said Steinberg.
"Machine learning always works best with real-world data. We have an algorithm that's able to detect the PFM-1 (butterfly mine) plastic landmine that's actually now being used and deployed in Ukraine," he added.
Evidence suggests that Russia is still deploying the same mines in Ukraine.
"So we're flying over, taking surveys of minefields, and then we're using machine learning to help us automate the detection afterward," said Steinberg.
The project is aimed to detect many types of cluster munitions and landmines.
“The goal would be to fly over with different sensors, either a thermal sensor, multispectral or visual input the imagery into the algorithm, and it'll output the coordinates of where different mines lay. And we'll also say what type of mine they are," Steinberg added.
According to the Landmine and Cluster Munition Monitor, mines killed or injured at least 7,073 people in 54 countries and areas in 2020. Many of the organizations working to remove these old munitions are non-profits with a fraction of the resources of the militaries that used them.
Scientists propose using dynamic soaring, a flying technique of seabirds, for much faster spacecraft propulsion that could let us reach faraway planets.