NIR hyperspectral images for identification of tagged gunshot residue: analyzing the collecting substrate

Gunshot residues (GSR) are particles produced during the discharge of a firearm. They are formed by the combustion products from both the primer and the propellant, as well as by some metallic particles arising from cartridges and firearm. GSR identification is a point of first interest in forensics, however its collection and identification is not a trivial task, especially when heavy metals-free ammunition (also called non toxic ammunition, NTA) is used. Aiming to shed a light on GSR analysis, our group had proposed to incorporate some luminescent taggants into ammunition. In this case, luminescent GSR (LGSR) are formed and a completely new range of possibilities for the collection and analysis is opened. One of these possibilities relates to the use of NIR hyperspectral images to distinguish among different taggants, as well as to provide information about the spatial distribution of LGSR. In this work, the potential of near infrared hypespectral images (HSI-NIR) associated with partial least squares discriminant analysis (PLS-DA) started to be evaluated for this purpose. The first step was the evaluation of LGSR collecting substrates.Five different substrates were tested: conductive double face carbon tape, ordinary adhesive tape, double face adhesive tape, teflon tape and glue pad. The usual process to GSR analysis uses the conductive double face carbon tape because it allows scanning electron microscopy observation. The other substrates were selected taking into account their adhesive properties and previous works concerning GSR in the literature. The LGSR samples were produced by incorporating the luminescent taggant∞[Eu(DPA)-(HDPA)] in the 9 mm non-toxic ammunitions (Clean Range® CBC).Shots were fired using GLOCK 17 pistols at indoor shooting range in the ballistics service of the NIC/BFP (National Institute of Criminalistics of the Brazilian Federal Police). After shooting, LGSR were collected on shooters’ hands, on cartridges and on firearms, totalizing 15 samples. HSI-NIR were acquired (928-2524 nm) using the camera SisuChema from Specim, with pixel size of 30 x 30 µm. The images were preprocessed using standard normal variate (SNV) and Savitzky-Golay smoothing filter (11 points window width and 2nd order polynomial). Images from each substrate (without LGSR) and from the luminescent taggant were also obtained in order to build PLS-DA models considering two classes: LGSR and substrate. The spectra of carbon tape is very noise with high absorbance values, being difficult the discrimination between classes. The double face adhesive tape showed a high degree of superposition with taggant spectrum, so it is not suitable for NIR analysis. The teflon tape, although showing small superposing with the taggant spectrum, was not effective in fixing the LGSR. The glue pad spectrum shows some windows where there is no superposing with the taggant spectrum, however glue pad texture’s is not uniform and focus problem was perceived. The PLS-DA models using LGSR samples collected on ordinary adhesive tape showed the best performance. This tape has uniform texture, adequate adhesive properties and a low degree of superposition of spectra. In the prediction images for the LGSR class, similar distributions with those obtained by UV irradiation were observed.