When trying to locate something hidden in a vast area, having a bird’s eye point of view can increase your chances of success. That is why drones have become critical tools in search and rescue missions. A pilot flies the drone through a prescribed grid, scanning the area below. The drone’s high definition and thermal camera can identify a person whether they are hidden among trees, rocks, in the water, almost anywhere. The same principle can be used to locate inanimate objects, though it requires a bit more programming to get a drone to recognize something like a meteorite.

This is exactly what Planetary Scientist in training Seamus Anderson decided to do. Seamus got his B.S in Physics from the University of Central Florida. From there he had the opportunity to intern at the NASA Goddard Space Flight Center in Greenbelt, MD. Currently, he is in Perth, Australia as a graduate student at Curtin University. As part of his graduate studies, Seamus is the lead researcher on a project using drones to locate meteorites through Curtin’s Desert Fireball Network (DFN). The DFN is a network of cameras that track meteorites entering Earth’s atmosphere over western and southern Australia.

When the autonomous cameras pick up a meteorite, a team from DFN is then sent out to try and locate it. However, finding them can be extremely difficult. It helps that the DFN’s cameras can narrow down the fall zone of a meteorite, but it can still take weeks to locate one of the small celestial rocks. Seamus proposed using drones, much like how a search and rescue team would, to scan the suspected fall zone for meteorites. The first step was for Seamus to develop an AI algorithm that would teach the drone to recognize what a meteorite looks like.

On April 1, 2021, the DFN tracked a meteorite entering Earth’s atmosphere over the Nullarbor Plain in Western Australia. The small meteorite landed somewhere within the vast desert like landscape in less than 3 seconds, according to data collected by 2 DFN cameras. So, Seamus had an idea of where to look for his treasure, a 5 square kilometer range of rock, brush, and debris covered land. “A camera-fitted drone flies over and collects images of the fall zone,” Seamus explained, “which are transferred to our field computer where an algorithm scans each image for meteorites and features that resemble them.”

Within 4 days of testing the drone system, Seamus found his “needle in a haystack”, a meteorite roughly the size of a small potato, next to a low bush. While by search and rescue standard, 4 days is unacceptable, for Seamus it was a huge success. “Meteorite searches usually involve a group of people walking over a large predicted impact area,” he said. “But our new method requires only about one tenth the amount of labor and time and has a much higher likely success rate, which is evident in the fact we located and recovered the meteorite within four days of being on site at Kybo Station.” Seamus went on to explain that the drone AI model he developed has implications beyond planetary science. “Other potential applications for our new approach using drones and artificial intelligence include wildlife management and conservation, as our model could be easily retrained to detect objects other than meteorites, such as plants and animals,” he said. Seamus’ novel drone study is just another reason to support the progression of drone technology.