PNNL

Abstract

This manuscript presents a method for categorizing gamma-ray spectra as benign or threatening. It is widely believed that the goal of segregating gamma-ray spectra into benign and threatening populations can achieved with fewer counts than are required for confident characterization of a spectrum’s isotopic composition, while still providing improvement over count-based algorithms. This has potentially important implications on the detection of radiological and nuclear threats, where decisions must be made from analysis of count-starved spectra that dominate the landscape of monitoring special nuclear material transport and lost-or-stolen source search. We report here the method of Spectral Comparison Ratios (SCRs) which is useful in the targeted detection of specific gamma-ray signatures or signature classes. SCRs discriminate between benign and target sources by comparing counts in broad, pre-defined energy bins that are pre-determined using statistical discrimination criteria. The integral component of the SCR algorithm is the location and interdependence of the energy bins, and we discuss the statistical methods used for choosing their locations along with the decision criteria that maximally separate targets from benign sources.