Multivariate analysis of standoff laser-induced breakdown spectroscopy spectra for classification of explosive-containing residues

Abstract

A technique being evaluated for standoff explosives detection is laser-induced breakdown spectroscopy (LIBS). LIBS is a real-time sensor technology that uses components that can be configured into a ruggedized standoff instrument. The U.S. Army Research Laboratory has been coupling standoff LIBS spectra with chemometrics for several years now in order to discriminate between explosives and nonexplosives. We have investigated the use of partial least squares discriminant analysis (PLS-DA) for explosives detection. We have extended our study of PLS-DA to more complex sample types, including binary mixtures, different types of explosives, and samples not included in the model. We demonstrate the importance of building the PLS-DA model by iteratively testing it against sample test sets. Independent test sets are used to test the robustness of the final model.