A GAN-based model upsamples low-statistics simulated muography images to look like 21 to 31 days of sampling, while a segmentation model identifies rebar and tendon ducts and, the authors claim, removes vertical smearing artifacts.
Cosmic rays for imaging cultural heritage objects
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
In cultural heritage conservation, it is increasingly common to rely on non-destructive imaging methods based on the absorption or scattering of photons ($X$ or $\gamma$ rays) or neutrons. However, physical and practical issues limit these techniques: their penetration depth may be insufficient for large and dense objects, they require transporting the objects of interest to dedicated laboratories, artificial radiation is hazardous and may induce activation in the material under study. Muons are elementary particles abundantly and freely produced in cosmic-ray interactions in the atmosphere. Their absorption and scattering in matter are characteristically dependent on the density and elemental composition of the material that they traverse, which offers the possibility of exploiting them for sub-surface remote imaging. This novel technique, nicknamed "muography", has been applied in use cases ranging from geophysics to archaeology to nuclear safety, but it has been so far under-explored for a vast category of cultural heritage objects that are relatively large (from decimeters to human size) and dense (stone, metals). The development of portable muon detectors makes muography particularly competitive in cases where the items to be analysed are not transportable, or set up in a confined environment. This document reviews the relevant literature, presents some exemplary use cases, and critically assesses the strengths and weaknesses of muography in this context.
fields
eess.IV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Muographic Image Upsampling with Machine Learning for Built Infrastructure Applications
A GAN-based model upsamples low-statistics simulated muography images to look like 21 to 31 days of sampling, while a segmentation model identifies rebar and tendon ducts and, the authors claim, removes vertical smearing artifacts.