LAUSR

Abstract In forensic examination of questioned documents a type of casework often encountered are the frauds that occur by mean of addition and adulteration of parts of text or numbers on document. The goal of this work is to evaluate the performance of hyperspectral images (HSI) in the near (NIR) and middle (MIR) regions, combined with the unsupervised pattern recognition techniques Principal Component Analysis (PCA) and Projection Pursuit (PP) for a rapid, reliable and non-destructive identification of document falsifications by means of alterations and additions. Blind tests were conducted for this purpose. Sixteen black ink pens from different brands, models and ink types were employed to prepare the samples in two different ways: (i) for initial discrimination and method validation, straight lines of approximately 2 cm long were produced in white paper; (ii) for blind testing, three collaborators used any of the sixteen pens available and prepared genuine or altered/added numbers (in total 30 samples) in white paper and in bank check paper. Overall, PP analysis showed better results than PCA to discriminate the 120 pairs of ink lines in white paper using HSI-MIR (97.5% and 87.5%, respectively). It is important to mention that the 10.0% of pairs that were not discriminated by PCA analysis were discriminated by PP, which highlights the importance of the combined use of the two chemometric techniques. HSI-NIR in combination with PCA and PP analysis was able to solve 76.7% and 83.3% of the blind testing samples, respectively. When HSI-MIR was used in a complementary way to HSI-NIR, discrimination of blind test samples increased to 90%. Therefore, HSI-NIR and HSI-MIR combined with PCA and PP show great discrimination potential and provide objective examination of suspected fraudulent documents.