Skip to Main Content
The scanning of cargo for radiological and nuclear material is vital in detecting the illicit trafficking of such materials. The deployment of technologies such as Radiation Portal Monitors (RPMs) has enabled screening for the presence of gamma and neutron emitting radionuclides. Although the development of radionuclide detection algorithms is vital to the development of RPMs, only a small amount of the work exists in the published literature. This paper describes the development of an anomalous signature detection algorithm based on Principal Component Analysis (PCA). PCA involves the eigen decomposition of the correlation matrix of a training data set. The distance of an unknown observed spectrum from Naturally Occurring Radioactive Materials (NORM), in a 14 dimensional space, was used to assess the algorithm performance. The PCA algorithm showed an excellent 'anomaly detection' performance for a number of threat sources including Special Nuclear Materials (SNM's). The PCA algorithm has also demonstrated an improved performance over that of a commercially available peak search algorithm. The discrimination of the SNM's sources, from the NORM, consistently improved with increased counts, which is not always true for peak search based algorithms. The algorithm also performed well in count starved spectra, which is of relevance to border security applications of RPMs.