MULTIVARIATE-ANALYSIS OF STATISTICALLY POOR EDXRD SPECTRA FOR THE DETECTION OF CONCEALED EXPLOSIVES

Energy dispersive x-ray diffraction (EDXRD) has been developed as a to ol for the detection of explosives in passenger baggage. The measured spectra result from the combined diffraction from each of the material s within a scattering volume. Multivariate regression was used to iden tify known components within very noisy data, permitting the rapid det ection of explosive materials in the presence of overlying media for s ecurity screening applications, Explosives can be positively identifie d in spectra containing as few as several hundred counts and the error associated with the prediction is consistent from statistically relia ble data (10(6) integrated counts) down to spectra containing in the r egion of 10(3) counts. This analysis can be employed in any situation where qualitative information is required from poor quality spectral d ata. (C) 1998 John Wiley & Sons, Ltd.