Machine Learning for Mini-EUSO Telescope Data Analysis

TL;DR

This paper demonstrates the effectiveness of machine learning, particularly neural networks, in classifying and recognizing various atmospheric phenomena and cosmic ray signals in data collected by the Mini-EUSO telescope onboard the ISS.

Contribution

The paper introduces ML-based methods for recognizing and classifying atmospheric and cosmic signals in Mini-EUSO data, showing that simple neural networks perform impressively.

Findings

Neural networks effectively classify atmospheric phenomena.

ML approaches distinguish meteors, space debris, and cosmic ray signals.

Simple neural networks achieve high performance in signal recognition.

Abstract

Neural networks as well as other methods of machine learning (ML) are known to be highly efficient in different classification tasks, including classification of images and videos. Mini- EUSO is a wide-field-of-view imaging telescope that operates onboard the International Space Station since 2019 collecting data on miscellaneous processes that take place in the atmosphere of Earth in the UV range. Here we briefly present our results on the development of ML-based approaches for recognition and classification of track-like signals in the Mini-EUSO data, among them meteors, space debris and signals the light curves and kinematics of which are similar to those expected from extensive air showers generated by ultra-high-energy cosmic rays. We show that even simple neural networks demonstrate impressive performance in solving these tasks.