Detection and classification of explosive substances in multi-spectral image sequences using linear subspace matching

Fast detection and analysis of dangerous substances from longer distances is highly desired in many security applications. Imaging Raman spectroscopy is a novel multi-spectral imaging technique designed for stand-off screening and detection of explosive substances. In this paper we present a method for detection and classification of explosive substances in multi-spectral image sequences from imaging Raman spectroscopy using linear subspace matching. Our approach uses limited spectral information and is computationally efficient, which enables fast screening of interesting areas. The performance of the method is evaluated on real stand-off measurements from a demonstrator system. We show that the method can detect and classify substances with high accuracy.