Time series evaluation of portal monitor data

Abstract

Radiation portal monitor (RPM) systems perform cargo screening at U.S. and other ports of entry. The data from these systems are routinely evaluated in regard to energy spectrum and total count rate above background. However, few analyses are performed on the time progression of data. In this work, the time series of data from an RPM system are evaluated for the presence of sources of interest by isolating the contribution of anomalous radiation. Source contributions are isolated by comparing the background spectra (or energy-windowed spectral information from low energy-resolution systems) to the spectrum at each successive time step. At every time in the data sequence, the total gross-count signal is turned into a “spectral distance” index using this method. This has the potential to dramatically reduce systematic fluctuations due to background attenuation by a vehicle (the so-called “shadow shielding” effect), and allow for time-series shape fitting for source size discrimination. The anomalous time series is reanalyzed for the presence of compact sources by using a wavelet filter function of similar size to the expected source profile. This may allow a dramatic reduction in gross-count alarm thresholds, leading to a corresponding sensitivity increase. This increase is shown by analysis of a number of real drive-through data sets taken at a U.S. port of entry. A set of isotopes of interest are injected into the data set, and the resultant “benign” and “injected” data sets are analyzed with gross-counting, spectral-ratio, and spatial algorithms. Spatial analysis alone was not sufficient to increase overall sensitivity given this data set, but both methods together gave a significant increase to detection performance.