SPATIAL RADIATION PROFILE CHARACTERIZATION FOR DETECTION OF THREAT-LIKE SOURCES

Radiation portal monitors (RPMs) are deployed at many ports to measure gamma and neutron radiation as traffic passes through generating a time series of measured counts. RPMs encounter a wide range of radiation sources from innocent naturally occurring radiative materials (NORM) to threat materials. At some locations the large number of NORM alarms caused by legitimate commercial goods creates an undue burden on responders to ascertain and adjudicate alarms. Often, a lengthy secondary inspection is required to determine the cause of the alarm thus slowing commerce and requiring increased manpower. Performing secondary inspections on all primary alarms is often not practical forcing operators to decide which primary alarms should undergo secondary inspection. This work proposes a method that could lead to improved hold – release decisions resulting in a reduction in the need for secondary inspections. RPMs with polyvinyl toluene gamma detectors only generate gross counts, that is, without any spectral information. For such detectors, and when the alarm rate is too high, affecting the stream of commerce, it is useful to have filters capable of discriminating between benign sources and potential threat sources. The spatial information content in measured time profiles could provide such a filter. This work investigates methods for spatial characterization of measured profiles and preliminary analysis with limited data shows potential benefit from using such a filter for reducing the number of required secondary inspections. The analysis uses metrics derived from measured profiles with set thresholds to obtain a probability of detection for a given probability of false alarms. Receiver operating characteristic (ROC) curves can be generated by varying the metric thresholds. The ROC curves are highly dependent on the datasets that are used to represent benign and threat sources. We will discuss the datasets used; describe the analysis methods, the profile characterization metrics, and the optimal values for metric thresholds. A sample ROC curve along with the corresponding metric thresholds will be presented. Operational assessment of this methodology falls outside of the scope since it requires stream of commerce data and realistic threat source datasets including realistic masking scenarios not currently available.