Scholarlink

<p>Cosmic-ray muon sources exhibit distinct scattering angle distributions when interacting with materials of different atomic numbers (<i>Z</i> values), facilitating the identification of various <i>Z</i>-class materials, particularly radioactive high-<i>Z</i> nuclear elements. Most traditional identification methods are based on complex statistical iterative reconstruction or simple trajectory approximation. Supervised machine learning methods offer some improvement but rely heavily on prior knowledge of the target materials, significantly limiting their practical applicability in detecting concealed materials. To the best of our knowledge, this is the first study to introduce transfer learning into muon tomography. We propose two lightweight neural network models for fine-tuning and adversarial transfer learning, utilizing muon scattering data of bare materials to predict the <i>Z</i>-class of materials coated by typical shieldings (e.g., aluminum or polyethylene), simulating practical scenarios such as cargo inspection and arms control. By introducing a novel inverse cumulative distribution-based sampling method, more accurate scattering angle distributions could be obtained from the data, leading to an improvement of nearly 4% in prediction accuracy compared with the traditional random sampling-based training. When applied to coated materials with limited labeled or even unlabeled muon tomography data, the proposed method achieved an overall prediction accuracy exceeding 96%, with high-<i>Z</i> materials reaching nearly 99%. The simulation results indicate that transfer learning improves the prediction accuracy by approximately 10% compared to direct prediction without transfer. This study demonstrates the effectiveness of transfer learning in overcoming the physical challenges associated with limited labeled/unlabeled data and highlights the promising potential of transfer learning in the field of muon tomography.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transfer learning empowers material Z classification with muon tomography

  • 期刊论文

  • 发表日期：

    2026年2月9日

查看全文
  • Hao-Chen Wang,
  • Zhao Zhang,
  • Pei Yu,
  • Yu-Xin Bao,
  • Jia-Jia Zhai,
  • Yu Xu,
  • Li Deng,
  • Sa Xiao,
  • Xue-Heng Zhang,
  • Yu-Hong Yu,
  • Wei-Bo He,
  • Liang-Wen Chen,
  • Yu Zhang,
  • Lei Yang,
  • Zhi-Yu Sun

摘要

Cosmic-ray muon sources exhibit distinct scattering angle distributions when interacting with materials of different atomic numbers (Z values), facilitating the identification of various Z-class materials, particularly radioactive high-Z nuclear elements. Most traditional identification methods are based on complex statistical iterative reconstruction or simple trajectory approximation. Supervised machine learning methods offer some improvement but rely heavily on prior knowledge of the target materials, significantly limiting their practical applicability in detecting concealed materials. To the best of our knowledge, this is the first study to introduce transfer learning into muon tomography. We propose two lightweight neural network models for fine-tuning and adversarial transfer learning, utilizing muon scattering data of bare materials to predict the Z-class of materials coated by typical shieldings (e.g., aluminum or polyethylene), simulating practical scenarios such as cargo inspection and arms control. By introducing a novel inverse cumulative distribution-based sampling method, more accurate scattering angle distributions could be obtained from the data, leading to an improvement of nearly 4% in prediction accuracy compared with the traditional random sampling-based training. When applied to coated materials with limited labeled or even unlabeled muon tomography data, the proposed method achieved an overall prediction accuracy exceeding 96%, with high-Z materials reaching nearly 99%. The simulation results indicate that transfer learning improves the prediction accuracy by approximately 10% compared to direct prediction without transfer. This study demonstrates the effectiveness of transfer learning in overcoming the physical challenges associated with limited labeled/unlabeled data and highlights the promising potential of transfer learning in the field of muon tomography.