The goal of this project was to profile the structural and asymptotic statistical properties of demonstrated data modeling, as well as to use state-of-the-art computational resources and modern methods of statistical analysis in making statistically well-founded assessments of the rarity of individualizing information for a latent fingerprint image.
The forensic science problem addressed in this project stems from the fact that the assessment of latent prints from crime scenes is based largely on human interpretation that typically claims to have zero error rates, which is scientifically implausible. The research led to the following features of latent fingerprint examination: 1) A Region of Interest (ROI) is specified in the latent image, and 2) The latent ROI is consistent with the quality areas of a particular exemplar image of interest. The research project used an objective measure of similarity applied to a large set of known non-mate fingerprint images randomly selected from a data base as a means for statistically estimating a data base Random Match Probability (RMP) for a specific latent image. The Objective Measure of Similarity (OMS) for one fingerprint image is a score that results from using data analysis in hierarchically narrowing tens of thousands of micro similarity scores to a single value. This project demonstrates how the data analyses create the conditions that lead to Objective Similarity Measurement from a large, randomly selected set of known no-mate fingerprint images that provide the data for a valid model to predict a fingerprint’s random Level 2 similarity to the latent image. By providing latent fingerprint examiners with an objective measure of similarity between an exemplar image and a latent print, together with an associated statistical random match error statement, constitutes a significant step in putting latent fingerprint examination on a scientific base. 31 figures