Preliminary study on using near-infrared spectroscopy at 1.6–2.4 µm for document examination

Abstract

We investigated whether near-infrared spectroscopy could be used to identify paper and colorants as a first step to applying the method to document examination. We verified whether samples that are difficult to differentiate by comparing their spectra with the naked eye could be distinguished by using principal component and discriminant statistical analyses. Eight types of printer paper and 10 type 1 colorants were used as experimental samples. A portable near-infrared spectrometer (microPHAZIR, Thermo Scientific) was used for the spectral measurement. Spectra were analyzed first by principal component analysis, and then by discriminant analysis. The classification model was evaluated by five-step cross-validation. All the papers were distinguished, and the black colorants were classified more accurately by first using all the principal components, and then using discriminant analysis. Our results show that near-infrared spectroscopy can be used to identify papers and black colorants. Even where the spectral shapes were similar, the secondary discriminant analysis identified the samples with high probability, confirming the effectiveness of secondary discriminant analysis.