Utilization of Artificial Intelligence for Meteor and Meteorite Recognition

Artificial Intelligence (AI) applications with its subfields are widely used in Astronomy and Space Sciences. As part of the meteorite center at the Sharjah Academy for Astronomy and Space Sciences (SAASST), we have developed two object detection algorithms that can recognize meteors and meteorites, respectively. The algorithms are based on the "You Only Look Once" (YOLOv3) deep learning architecture which are constructed using Convolutional Neural Networks (CNN). The dataset of meteors is built using images observed and recorded by the United Arab Emirates (UAE) Meteor Monitoring Network (MMN) stations, which includes meteors and other noise objects such as airplanes and insects that are often mistakenly detected by the cameras. Conventionally, recorded files are manually filtered daily. However, with AI, the process can be further facilitated. Our dataset also includes individual images of meteorites from the collection at SAASST and rocks collected from neighboring areas. Since meteorites and rocks could be very confusing to an amateur meteorite hunter, the developed algorithm extracts and learns features of these objects and is able to distinguish between them based on their external appearance. Results have shown that the meteor algorithm can detect faint and bright meteors with an accuracy ranging from 80%-90%. As for meteorites, the algorithm was better at recognizing iron meteorites compared to stony meteorites. Such study opens potential areas for computer vision directed to studying small bodies of the solar system from a different perspective.