A meteorite has been successfully recovered using drones and a machine-learning algorithm for the first time. The development marks a step forward in improving the process of searching Earth鈥檚 surface for extraterrestrial rocks.
Typically, when a meteorite falls to Earth, it takes a team of five or more people to visually search the ground around the fall zone to find it. While the process can be successful, it is costly and time-consuming.
at Curtin University in Australia and his colleagues used a drone to take aerial images of a dry plain in Western Australia where a fireball meteor was spotted in the sky in December 2021. These images were then inspected by a machine-learning algorithm that had been trained to identify photos of actual meteorites in desert conditions. The algorithm had been improved further through training on an archive of photographs of meteorites.
Advertisement
The algorithm picked out potential meteorites from the images, which the team then manually checked to remove any false positives. The team then sent out another drone to capture more photographs of the most promising meteorite candidates, and these images were inspected, too, to discount more false positives.
Finally, the team visited the four remaining candidates in person and recovered a 5-centimetre-long meteorite weighing 70 grams. The process from the reported meteorite fall to the recovery of the tiny rock took just four days. Although the meteorite hasn鈥檛 been classified yet, the authors say it resembles common stony meteorites known as chondrites.
The work, which was presented at the on 8 March, may help to streamline future meteorite hunting. The technology could also be used to locate meteorites in more remote locations, such as in Antarctica.
鈥淎s the technology continues to improve, it will decrease the time it takes to recover meteorites from future falls,鈥� says at the University of California, Davis, who trialled similar technology last year. 鈥淭his is especially important for smaller meteorite falls that would not attract many meteorite hunters.鈥�
鈥淚 am super-excited by the prospect of this and it鈥檚 great that they were able to collect a meteorite within four days of a fall using machine learning,鈥� says at the University of Glasgow, UK. 鈥淚t probably helps a lot that it fell in a desert environment, as it鈥檚 a high contrast environment, so probably easier for the algorithm to spot a black rock against the desert sand. I can鈥檛 see this being quite so easily applied to fireballs in lower contrast environments, though, because they would be much trickier to spot.鈥�
