The primary aim for this research is to capture, analyse and interpret anatomical variation in images of the human hand through the combinatorial lens of a best-fit multimodal biometric. Success will occur if no two hands can be found that are identical, implying ‘uniqueness’.Key Objectives:
This research is being used primarily for the comparison of images between suspect and perpetrator in cases of child sexual abuse.
The first database of images will involve approximately 500 participants who will have their hands photographed under natural and near infrared (NIR) light. This database forms the basis for the development of algorithms that will be developed and refined to allow the automatic extraction of features of anatomical interest through machine learning. These algorithms will firstly be tested on the particpants in a second database. This will comprise around 5000 participants of the general public who will take photographs of their own hands using their mobile phones. The images will be anonymised and used as a test bed for the derived algorithms to identify the hands of an individual from their anatomical features.Following black and white box testing, the algorithms will be tested on police datasets of images of child sexual abuse. The aim is to identify whether a prepetrator occurs more than once in the international datasets of offending images.A third dataset of 3d scans will be analysed to determine how much anatomical information is lost as a result of movement of the hand.