Pen Ink Library: An interactive database of writing instruments based on Vis-NIR reflection spectra and optical...

Many tasks in forensic examination of handwritten documents require classification of writing instruments that have ink of similar properties as the ink found on a questioned document. In this paper, we propose a new methodology for non-destructive identification of inks based on optical properties and reflectance spectra of the ink, measured from handwriting strokes. Building on this methodology, we developed an interactive database that we call the "Pen Ink Library", which lists 718 various writing instruments and enables systematic comparison and semi-automatic search of writing instruments, using the measured characteristics of their ink. To highlight the significance and applicability of the database, we additionally exploit the large amounts of collected measurements to design computer-based data analysis methods for classification and comparative analysis of ink samples. For validation of the semi-automatic search functionality of the Pen Ink Library we performed a series of blind tests using twenty randomly selected writing instruments. Here, an instrument with the same brand and model was found in nine cases, and an instrument with a different brand and model, but with identical spectrum and optical parameters, was found in five cases. Cross-validation of the computer-based data analysis methods on the measurements from the database yielded above 90% accuracy of the classification method and 5.3% to 12.7% error rate of the comparative analysis method.