Automated isotope identification algorithms for low-resolution gamma spectrometers

Handheld radio-isotope identifiers (RIIDs) are widely deployed for nuclear security applications, but these detectors generally have poor isotope identification performance. Most of these deployed detectors use low-resolution NaI scintillators due to their low cost and good efficiency. Alternative detection hardware could be used to improve performance, but better detectors are generally cost-prohibitive for wide deployment in this mission space. However, a trained spectroscopist can use these low-resolution detectors to make much more accurate identifications than the RIIDs produce. For this reason, it has been suggested that these RIIDs could be significantly improved by changing the onboard identification algorithms. To this end, a peak-based Bayesian classifier has been developed to perform automated isotope identification. This algorithm was constructed to mimic the manual identification that a spectroscopist would perform. This approach can handle challenges such as detector calibration drift and unknown shielding scenarios, is capable of identifying mixed radiation sources, and is computationally inexpensive enough to be feasible for deployment on handheld RIID systems. A method for easily generating isotope libraries that are coupled to the detector and to the feature extraction algorithm is presented as well. This method is demonstrated on a broad variety of gamma-ray spectra, ranging from small calibration sources ($<$ 1 μCi) to Category I quantities of special nuclear material. This algorithm is also benchmarked against the ANSI N42.34-2006 Standard for Handheld Identifiers as a part of the Department of Homeland Security Algorithm Improvement Program.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Nuclear, Plasma, Radiolgc Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2017

Author dc:creator

• Stinnett, Jacob Benjamin

Contributors dc:contributor

• Sullivan, Clair J.
• Stubbins, James F.
• Ruzic, David N.
• Varshney, Lav R.

• Gamma-ray spectroscopy
• Bayesian classifiers
• Isotope identification
• Wavelets

Statement dc:rights

• Copyright 2016 Jacob Benjamin Stinnett

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/95304

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/95304