Neural Network Technique Based on Hyperspectral Imaging for Determination of Blood Stains in Tissues Washed with Bleach

The recognition of cleaned blood holds significant importance within forensic inquiries, aiding in the identification and reconstitution of crime scenes. Traditional techniques for detecting cleaned blood rely on chemical assays or the utilization of Luminol, both of which may potentially compromise and modify the crime scene integrity. Nowedays, hyperspectral system is widely used in the forensic field to investigate in a non destructive way a crime scene. The present work focuses on advancing an HSI-guided methodology aimed at identifying whether bleach has ever been used in a white tissue and, in parallel, whether blood has ever been deposited in a tissue that appears clean. This goal is achieved by means a two Multi-Layer Perceptron (MLP) networks able to perform a binary classification from the collected spectra; in the first model the aim is to detect the presence of bleach in the tissue while in the second one the objective is to investigate the presence of blood in cleaned tissues. The developed model achieves a remarkable accuracy of 99.9% in identifying the reflectance spectra of tissues where blood has been removed and in tissues where bleach has been dropped.