Anomaly Metrics to Differentiate Threat Sources from Benign Sources in Primary Vehicle Screening

摘要

Discrimination of benign sources from threat sources at Port of Entries (POE) is of a great importance in efficient screening of cargo and vehicles using Radiation Portal Monitors (RPM). Currently RPMs ability to distinguish these radiological sources is seriously hampered by the energy resolution of the deployed RPMs. As naturally occurring radioactive materials (NORM) are ubiquitous in commerce, false alarms are problematic as they require additional resources in secondary inspection in addition to impacts on commerce. To increase the sensitivity of such detection systems without increasing false alarm rates, alarm metrics need to incorporate the ability to distinguish benign and threat sources. Principal component analysis (PCA) and clustering technique were implemented in the present study. Such techniques were investigated for their potential to lower false alarm rates and/or increase sensitivity to weaker threat sources without loss of specificity. Results of the investigation demonstrated improved sensitivity and specificity in discriminating benign sources from threat sources.