Point-Source Detection Using Gamma-Ray Spectra in Radiation-Portal Monitoring

Abstract

Problem: Radiation detection systems are deployed at U.S. borders to guard against illicit entry of radioactive material. Unfortunately, nuisance alarms due to naturally occurring radioactive material (NORM) reduce detection probabilities for threat sources. This paper evaluates to what extent gamma counts for a range of energies (i.e., a gamma spectrum) deployed in primary screening can detect threat-point sources, with attention to background-suppression effects.

Approach: It is assumed that most NORM sources are more spatially distributed than a point source and that most threat sources are distributed more like a point source. Therefore, we seek a high-alarm probability for vehicle profiles that exhibit a point-like shift in spectral shape. The proposed approach uses variation of the gamma spectral shape over time as a moving vehicle carrying a point source is screened.

Results: The paper uses data from approximately 2000 real vehicle screenings to develop alarm thresholds and to evaluate detection probabilities of injected threat sources from both NORM and threat isotopes. It is shown that transforming the raw gamma spectra is advantageous and that using the Mahalanobis distance to detect the point-like shift in spectral shape is superior to a spectral distance used in the literature.

• Background Suppression
• Mahalanobis Distance
• Scan Statistic
• Spectral Distance
• Time Series

Additional information

Notes on contributors

Tom Burr

Dr. Burr is a Scientist in the Statistical Sciences Group. His email is [email protected].

Michael S. Hamada

Dr. Hamada is a Scientist in the Statistical Sciences Group. He is an ASQ member. His email is [email protected].

Kary Myers

Dr. Myers is a Scientist in the Statistical Sciences Group. Her email is [email protected].

Misha Skurikhin

Mr. Skurikhin is a Student Intern in the Statistical Sciences Group. His email is [email protected].