Análise discriminante em química forense: aplicações em documentoscopia

Abstract

The analysis of inks in paper represents an important area in forensic chemistry, since it can be used to identify possible falsifications and undue changes in a document. However, most of the instrumental methods are unable to perform a nondestructive analysis, which represent a serious limitation in document examination. This work proposes the application of partial least squares discriminante analysis (PLS-DA) method applied in digital images obtained from smartphones (iPhone 5s e Motorola Moto G II generation) for nondestructive identification of blue pen inks (Ballpoint, Rollerball, Gel and Felt-Tip) and crossed dash analysis in a document that was printed after writing and in a document that was written after printed. PLS-DA models were developed for each ink from different pen and showed sensitivity equal to 1.0. Specificity was better in the model for Ballpoint and Gel inks. The robustness of these models, measured from images obtained with different paperweight and from a Motorola phone, showed that Felt-tip ink produces the most robust model. The models for crossed dash analysis presented sensitivity and specificity equal to 1.0. In this case, the gel ink model was the only one that showed robustness when the printer was changed.