Meteorite Hunting Using Deep Learning and UAVs

In this paper, we present an automated meteorite detection system that employs an autonomous Unmanned Aerial Vehicle (UAV). It is programmed to recognize and locate meteorites using machine learning. In this design, the UAV carries a Single Board Computer (SBC), through which realtime data processing of the live video feed from an optical camera is performed. The integration of a GPS module into the system enables localization of the detected meteorites. The onboard processing, carried out by the SBC, involves analyzing each frame from the video and running it through our deep learning model. This model is trained on images of different types of commonly found meteorites, via Transfer Learning performed on Convolutional Neural Networks (CNN). The results were promising, with an accuracy of approximately 90%. Future enhancements, including the addition of an infrared sensor to capture thermal images, are also discussed.