Detection of nuclear sources in search survey using dynamic quantum clustering of gamma-ray spectral data

Abstract

In a search scenario, nuclear background spectra are continuously measured in short acquisition intervals with a mobile detector-spectrometer. Detecting sources from measured data is difficult because of low signal to noise ratio (S/N) of spectra, large and highly varying background due to naturally occurring radioactive material (NORM), and line broadening due to limited spectral resolution of nuclear detector. We have invented a method for detection of sources using clustering of spectral data. Our method takes advantage of the physical fact that a source not only produces counts in the region of its spectral emission, but also has the effect on the entire detector spectrum via Compton continuum. This allows characterizing the low S/N spectrum without distinct isotopic lines using multiple data features. We have shown that noisy spectra with low S/N can be grouped by overall spectral shape similarity using a data clustering technique called Dynamic Quantum Clustering (DQC). The spectra in the same cluster can then be averaged to enhance S/N of the isotopic spectral line. This would allow for increased accuracy of isotopic identification and lower false alarm rate. Our method was validated in a proof-ofprinciple study using a data set of spectra measured in one-second intervals with Sodium Iodide detector. The data set consisted of over 7000 spectra obtained in urban background measurements, and approximately 70 measurements of Cs-137 and Co-60 sources. Using DQC analysis, we have observed that all spectra containing Cs-137 and Co-60 signal cluster away from the background.

Survey of Machine Learning Approaches in Radiation Data Analytics Pertained to Nuclear Security

“…To that end, the work in [58] employed DQC for anomaly detection in gamma-ray spectral data. In particular, the DQC approach that has been taken is to cluster the data to detect or not the presence of two nuclides, and more specifically Cs-137 and Co-60, i.e., in other words, DQC was looking for specific anomalous data related to those two nuclides.…”

Detection of Anomalies in Environmental Gamma Radiation Background with Hopfield Artificial Neural Network - Consortium on Nuclear Security Technologies (CONNECT) Q3 Report

“…Consortium on Nuclear Security Technologies (CONNECT) Q3 Report prepared by Luis Valdez 1,2 , Alexander Heifetz 1…”

Preliminary Assessment of Qiskit Quantum Simulator Capabilities for Development of Quantum Hopfield Neural Network for Anomaly Detection Applications (Q4 Report)

“…In a typical scenario, a mobile detector-spectrometer continuously measures gamma radiation spectra in short, e.g., one-second, signal acquisition intervals. Detecting sources from data measured in a search scenario is difficult due to the highly varying background because of naturally occurring radioactive material (NORM), and low signal-to-noise ratio (S/N) of spectral signal measured during one-second acquisition intervals [1][2][3]. In prior work, we developd a Hopfield Artificial Neural Networks (HANN) to detect a weak signal anomaly hidden among the highly fluctuating background spectra [4].…”