Automatic Detection if Explosives Using Laser-Based Standoff Spectroscopy

Kurzfassung

Real-time detection of explosive residues on various surfaces is of great importance in civil security applications. Laser-based spectroscopic techniques such as mid infrared (MIR) and Raman spectroscopy have great potential for remote trace detection of explosives. However, up to now, reliable substrate-independent data analysis remains a major challenge, as the substrate signatures often dominate the spectral response. Furthermore, especially for MIR spectroscopy, the backscattered signal strongly depends on the surface roughness, reflectivity and angle of incidence. Thus, machine-learning models frequently fail to generalize to previously unseen materials. Here, we present our results for reliable, substrate-independent detection algorithms for explosives such as TNT, RDX and PETN on fabric surfaces. Two distinct methodologies are developed for the two spectroscopy methods, namely UV Raman and MIR spectroscopy. The resulting detection performance for the two independent techniques are compared and the limitations and advantages of the methods discussed. Our developments represent an important step toward real-time detection of explosives in security applications.