Statistical properties of an algorithm used for illicit substance detection by fast-neutron transmission

摘要

A least-squares algorithm developed for analysis of fast-neutron transmission data resulting from non-destructive interrogation of sealed luggage and containers is subjected to a probabilistic interpretation. The approach is to convert knowledge of uncertainties in the derived areal elemental densities, as provided by this algorithm, into probability information that can be used to judge whether an interrogated object is either benign or potentially contains an illicit substance that should be investigated further. Two approaches are considered in this paper. One involves integration of a normalized probability density function associated with the least-squares solution. The other tests this solution against a hypothesis that the interrogated object indeed contains illicit material. This is accomplished by an application of the F-distribution from statistics. These two methods of data interpretation are applied to specific sets of neutron transmission results produced by Monte Carlo simulation.