Utilising Dynamic Time Warping in the Assessment of Alarms from Radiation Portal Monitors - An Exploratory Study

Border monitoring systems focused on detecting the illicit trafficking of nuclear and radioactive materials are now widely deployed at airports, vehicle crossings, maritime facilities and other strategic points. A key component of these systems are radiation portal monitors, typically used for the large scale initial scanning of goods, vehicles and people entering, leaving or transiting a particular location. To be effective radiation portal monitors (RPM) must be able to distinguish potential threat materials from legitimate radioactive shipments and naturally occurring radioactive materials that are regularly transported through these environments. This is a challenging task as, in order to avoid significant disruption to supply chains and immigration flows, alarming occupancies must be rapidly assessed, with potential threat materials distinguished from legitimate radioactive goods. This paper explores whether the data science technique of Dynamic Time Warping (DTW), could be used to support the assessment of initial alarms from radiation portal monitors. It examines how DTW can be used to nonlinearly warp RPM profiles of different spatial lengths so they can be compared and their relative similarity calculated. Clustering is then used to group alarming occupancies, demonstrating that this approach can preferentially cluster similar commodity types. Consequently, this approach if integrated into the alarm assessment process could provide new automated insights into the cause of initial RPM alarms.