Many of the mines and bombs dropped by the United States on Cambodia during the Vietnam War remain unexploded, according tomedia, a huge problem for those tasked with bomb removal. In response, scientists have developed a new artificial intelligence (AI) tool that can make things a lot easier.
It is understood that it can accurately detect the crater through satellite imagery to help reveal the possible location of the undetonated device.
The study was carried out by scientists from Ohio State University, who began with a satellite image showing a 100-square-kilometer area that was targeted by the 1973 U.S. carpet bombing. The computer algorithms the team first began using to detect craters on satellites and planets, but the researchers trained them to know the nuances of craters and craters caused by bombs.
This means taking into account the shape, color, texture, and size of the crater. This work is crucial because grass, shrubs and years of erosion alter the surface of the bomb pit.
At first, machine learning algorithms were able to detect 157 of the 177 real craters with an accuracy rate of 89 percent, while at the same time it identified 1,142 fake craters. In further studies, the system excluded 96 per cent of false craters, but a decrease in accuracy but only a slight decline — 152 of the 177 bomb pits still detected, with an accuracy rate of 86 per cent.
The team says its machine learning algorithm provides more than 160 percent of true bomb detection. This will be a powerful tool for helping to find and remove bombs that have not yet detonated.
Using declassified military data, the researchers calculated that a total of 3,205 carpet bombs had been dropped in the areas they studied. Once they know the number and location of the explosions, they can begin to work out how many did not explode and where they might be. They spend, and research shows that between 1,400 and 1,600 carpet bombs in the area have yet to explode.
It is understood that much of the land the team focused on in the study was used for agriculture, which meant that people working in the area were at serious risk. In Cambodia, one person is injured every week by an undetonated device, and more than 64,000 people have been injured or killed since the bombing of the area in 1973, the researchers said.
“The process of mine clearance is expensive and time-consuming, but our model can help identify the most vulnerable areas where mine clearance should be pioneered,” said Erin Lin, an assistant professor of political science at Ohio State University. “
The study was published in PLOS One.