Speaker: Joseph Zemmels, PhD Candidate, Department of Statistics, Iowa State University

Title: Automatic Scoring of Cartridge Case Impression Evidence

Abstract: Forensic examinations attempt to solve the binary classification problem of whether two pieces of evidence originated from the same source. For example, a cartridge case found at a crime scene may be compared to a cartridge case fired from a suspect’s firearm. Automatic comparison algorithms have grown in prevalence in a number of forensic disciplines following the reports from National Research Council (2009) and President’s Council of Advisors on Science & Technology (2016). We introduce the Automatic Cartridge Evidence Scoring (ACES) algorithm that uses surface impressions left by a firearm to compare three-dimensional topographical scans of cartridge cases. The ACES algorithm pre-processes the scans, extracts numeric features, and returns a similarity score indicating whether two cartridge cases were fired from the same firearm. We train and test the ACES algorithm using scans taken at the Roy J Carver High Resolution Microscopy Facility of cartridge cases collected by Baldwin et al. (2014). The performance of ACES compares favorably to several other methods, such as logistic regressions on smaller feature sets, random forests, and other predominant scoring methods. We provide a user-friendly interface for the algorithm in an interactive web application called "cartridgeInvestigatR."