标题： 非法货物探测的μ子成像：基于仿真的研究 显示英文标题

标题： Muon Imaging for Illicit Cargo Detection: A Simulation-Based Study

摘要： 本研究基于使用GEANT4工具包进行的详细模拟，评估了μ介子断层扫描作为一种非侵入性技术在探测货物中隐藏的违禁毒品方面的潜力。 结合μ介子散射和吸收数据的综合分析被用于增强材料分辨能力，重点研究了涉及可卡因藏匿于合法货物中的现实走私场景。 提出了一种两阶段检查协议以平衡检测速度与分辨率。 在第一阶段，约60秒的快速扫描用于识别异常的散射和吸收率，而无需进行完整的断层重建。 快速扫描数据的受试者工作特征（ROC）分析显示，随机森林分类器达到了曲线下面积（AUC）为0.9969，而多元正态似然模型达到AUC为0.9977，两者均表明对无害货物和走私违禁品之间有极佳的区分能力。 当发现异常时，会启动一次持续约30分钟的扩展扫描，以实现高分辨率三维成像，从而准确定位和识别隐藏的物质。 模拟结果显示，在快速扫描阶段，如果探测器的空间分辨率为1毫米（FWHM），那么可以在大约3倍标准差统计显著性水平上探测到隐藏的可卡因等违禁品。 在扩展扫描中，通过应用诸如DBSCAN等聚类算法于断层重建，成功地可视化并自动识别出藏匿在香蕉箱内的可卡因包裹。 这些发现证实了宇宙射线μ介子断层扫描作为一种被动、安全且有效的方法，在实际货物检查应用中用于探测违禁品的可行性。 显示英文摘要

摘要： This study evaluates the potential of muon tomography as a non-invasive technique for detecting concealed illicit drugs in cargo, based on detailed simulations performed using the GEANT4 toolkit. A combined analysis of muon scattering and absorption data was employed to enhance material discrimination, with a focus on realistic smuggling scenarios involving cocaine hidden within legitimate cargo. A two-stage inspection protocol is proposed to balance detection speed and resolution. In the first stage, a rapid scan lasting ~ 60 seconds is used to identify anomalous scattering and absorption rates, without requiring full tomographic reconstruction. Receiver Operating Characteristic (ROC) analysis of rapid scan data revealed that the Random Forest classifier achieved an area under the curve (AUC) of 0.9969, while the multivariate normal likelihood model attained an AUC of 0.9977, both indicating excellent discrimination between benign cargo and smuggled contraband. Upon detection of anomalies, an extended scan ~30 minutes is initiated to enable high-resolution three-dimensional imaging for accurate localization and identification of hidden materials. Simulation results demonstrate that, with a detector spatial resolution of 1~mm (FWHM), concealed contraband such as cocaine can be detected with approximately 3 sigma statistical significance during the rapid scan phase. In extended scans, cocaine packages concealed within banana boxes were successfully visualized and automatically identified using clustering algorithms such as DBSCAN applied to the tomographic reconstruction. These findings confirm the feasibility of cosmic-ray muon tomography as a passive, safe, and effective approach for contraband detection in real-world cargo inspection applications.