Detection of Nuclear Sources in Search Applications using Dynamic Quantum Clustering of Spectral Data
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-of principle 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.