Nondestructive identification of blue pen inks for documentoscopy purpose using iPhone and digital image analysis including an approach for interval confidence estimation in PLS-DA models validation

Abstract

Documentoscopy is a field of forensic science that studies documents related to crime investigations. In this sense, the evaluation of pen inks used to produce these documents can help in the counterfeits identification, and the analysis of inks in paper is an important area in documentoscopy. In order to develop a methodology that allows the nondestructive accurate discrimination of pen inks types (gel, rollerball, felt tip and ballpoint pens) partial least squares for discriminant analysis (PLS-DA) was applied for the digital images obtained from an iPhone. The proposed method was tested for discrimination between 42 blue pen inks of different types and brands and validated by independent validation (or test) samples. The interval confidences estimation were considered in the validation step. Evaluations about the effect of paper characteristics and influence of the utilization of a smartphone from another brand were described. The digital images coupled with PLS-DA model can be considered robust for felt-tip pen type, since the model for this pen type did not suffer influence of the paper grammage and smartphone brand. Therefore, the proposed methodology may be considered a relevant contribution to the forensic analysis of questioned documents, since it provides a simple, fast, nondestructive and efficient way to directly analyze blue pen inks.