Developing a new dataset and method for the interpretation of firearms evidence

Abstract

Firearms evidence has long been utilized in investigations and in the courtroom. Emerging technologies have made the examination of firearms evidence, namely expended cartridge cases, an easier process. Recent technologies enable the use of computer algorithms to provide quantitative comparisons. However, many of the cartridge case datasets used to develop these algorithms are limited in number and scope, in particular insufficient accounting for variability expected to occur within markings produced by a single firearm. The goal of this research was to develop a new dataset consisting of 31 HiPoint C9 firearms with 17 cartridge cases fired from each. These cartridges were scanned using the Cadre Forensic TopMatch 3D instrument and assigned similarity scores by the TopMatch 3D algorithm. These similarity scores were then analyzed using both kernel density estimation and generalized linear mixed models. Kernel density estimation had difficulties due to the low distribution of scores among firearms tested. GLMM, however, offered a means by which a percentage probability of same- or different-source conclusion could be assigned to similarity scores of a model of firearm. This method has the potential to be expanded to other firearms and offer some quantification to the field of forensic firearms analysis.