Background suppression with machine learning in volcano muography

Abstract

In this work, a machine learning algorithm, specifically a deep neural network, is introduced to mitigate background interference in muography applications, predominantly aimed at volcano imaging. The discussed detector system is engineered to filter out low-energy background by incorporating up to five lead absorber layers interspersed among eight detectors. This intricate system underwent a Monte-Carlo (Geant4) simulation to furnish training samples for the machine learning algorithm. It's demonstrated that the devised deep neural network substantially outperforms the traditional tracking algorithm in suppressing low-energy background, thereby rendering significant enhancement via machine learning supplementation.