Semester

Summer

Date of Graduation

2019

Document Type

Thesis

Degree Type

MS

College

Eberly College of Arts and Sciences

Department

Forensic and Investigative Science

Committee Chair

Keith B. Morris

Committee Co-Chair

Jacqueline Speir

Committee Member

Jacqueline Speir

Committee Member

Theunis Brits

Abstract

Firearms are one of the most commonly used weapons by violent criminals throughout the United States. In crimes involving the discharge of a firearm, residues from the combustion products are deposited on the shooter and in the firing direction. Some of the residues left behind are unburnt and partially burnt particles of the propellant. The most common propellant used in modern firearms is smokeless powder. The identification of particles as smokeless powder and the determination of the brand of the powder are two ways a forensic examiner may be able to utilize smokeless powder evidence. The determination of the brand of smokeless powder can provide investigators with investigative leads or as associative evidence when the same gunpowder is found in the possession of a suspect.

This study was broken up into two parts: the classification of unburnt smokeless powders by shape using physical measurements and image analysis; and the identification of the brand of smokeless powder from the color analysis of partially burnt powder grains recovered after the discharge of a firearm. The first part of the study utilized the open source software, ImageJ, to quickly and objectively record eight physical measurements of unburnt smokeless powder grains. The second part of the study used microspectrophotometry and the Commission Internationale de l’Eclairage 1976 L, a, b (CIELAB) color coordinate system to objectively analyze the color of partially burnt smokeless powder grains. The performance of both methods were analyzed using multivariate classification techniques with multinomial regression being used on the unburnt powder data and linear discriminant analysis (LDA) being performed on the partially burnt powder data. The second part of the study also used 6 blind samples to test the method. Additionally, a database was constructed for the partially burnt data using Microsoft® Access® to create a list of potential sources in the instance that a single powder brand could not be identified.

The shape classification of the unburnt powder grains resulted in an 8.3% misclassification rate when 70% of the data was used for training and 30% was used for testing. For the partially burnt powders, a leave-one-out cross validation during the LDA analysis among all of the smokeless powder brands resulted in a misclassification rate of 22.7%. Four of the 6 blind samples were classified to the correct smokeless powder brand using LDA. The correct powder brands of the 6 blind samples also appeared in the list of results when searching the database. The two methods investigated during this research project provide two objective classification methods to be used by a forensic examiner when smokeless powder evidence is presented.

Embargo Reason

Publication Pending

Download

Share

COinS