Muon tomography imaging algorithms for nuclear threat detection inside large volume containers with the Muon Portal detector

TL;DR

This paper evaluates various statistical algorithms, including autocorrelation, clustering, and log-likelihood methods, for reconstructing 3D images in muon tomography to detect nuclear threats inside large containers, using detailed simulations.

Contribution

It introduces and compares multiple reconstruction algorithms for muon tomography, highlighting their relative effectiveness in detecting high-Z objects within containers.

Findings

Autocorrelation and clustering algorithms show promising results.

Log-likelihood iterative method improves image accuracy.

Algorithms tested with realistic Geant4 simulations.

Abstract

Muon tomographic visualization techniques try to reconstruct a 3D image as close as possible to the real localization of the objects being probed. Statistical algorithms under test for the reconstruction of muon tomographic images in the Muon Portal Project are here discussed. Autocorrelation analysis and clustering algorithms have been employed within the context of methods based on the Point Of Closest Approach (POCA) reconstruction tool. An iterative method based on the log-likelihood approach was also implemented. Relative merits of all such methods are discussed, with reference to full Geant4 simulations of different scenarios, incorporating medium and high-Z objects inside a container.