Evaluation of fingerprint aging using near infrared hyperspectral imaging with partial least squares-discriminant analysis

Fingerprints are forensic evidence commonly found at crime scenes. In addition to being widely used for human identification purposes, they can also be used for dating, contributing information to criminal investigations. Despite the enormous gain that can be obtained with this type of approach, there is no reliable analytical method to estimate the age of a fingerprint. Aging is a complex and uncontrolled process, and changes are affected by factors such as time and environmental conditions. This study sought to evaluate the possibility of dating fingerprints using near-infrared hyperspectral images (NIR-HSI) using the chemometric tool Discriminant Analysis by Partial Least Squares (PLS-DA). Fingerprints from 6 volunteers were analyzed over an 8-week aging period, with this period divided into distinct time intervals for further classification. This resulted in three PLS-DA models that were compared: the first one with 8 classes, the second one with 4 classes and the last with only 2 classes. In the first model, 5.56 % of the prediction samples were correctly classified, while in the second model, a success rate of 30.56 % was obtained and the third model achieved a 63.89 % success rate. This highlights the difficulty of creating an effective fingerprint dating model, especially with a narrow period of accuracy.