Drone: About 500 meteorites fall to Earth each year, as estimated by planetary scientists, but only 2% of them are located. Trying to improve the numbers, researchers at the University of California at Davis, in the United States, are using drones with artificial intelligence (AI) to hunt these space rocks.
The automated search system uses a gridded virtual grid to establish the search area, based on where the fragments fall. From then on, the equipment records a series of photos within this field, according to the study published in June in the journal Meteoritics & Planetary Science.
In the next step, the algorithm analyzes the images taken by the autonomous drone, trying to identify potential meteorites. The AI has been trained with thousands of photos of recovered space stones, helping to improve its detection capability by differentiating this material from rocks of terrestrial origin.
The recovery of small pieces of meteorites is usually complicated, as they spread over large areas when they break, and can fall on trees, plants and places of difficult access. Their study provides important information regarding the formation of the Solar System.
In search of evolution
Autonomous meteorite hunting drones were first tested in 2019 in the Walker Lake region of Nevada (USA). According to planetary scientist Robert Citron, coordinator of the project, the concept revealed a series of false positives initially, but it has evolved.
In subsequent tests, Citron and his team managed to minimize the number of common rocks flagged as false detections, obtaining better results than the initial flight. And with the use of higher resolution cameras, the trend is for the system to be even more efficient.
“Fortunately, with each field test, we gain more data that we can incorporate into the pool and use to train the object detection network, improving accuracy,” Citron told Universe Today.