Date of Award

3-2025

Document Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Department of Electrical and Computer Engineering

First Advisor

Brett J. Borghetti, PhD

Abstract

The classification of uranium particles from scanning electron microscopy (SEM) imagery is critical to nuclear forensics, but has traditionally relied solely on skilled analysts whose classification accuracy and procedures may vary widely. Existing morphology lexicology [1] provides standardization guidelines to aid analysts but cannot fully address analyst variability. Using a dataset of 1,906 SEM images across 13 unevenly distributed particle classes and 73 magnification levels, final accuracy between statistical and deep learning methods were compared to find the best classification techniques. Ultimately, the deep learning model achieved an impressive 82% accuracy (80% balanced accuracy) on a withheld test set. This represents a 41% increase in accuracy (39% balanced accuracy increase) over statistical methods and a 72.3% improvement over a baseline naive model. The 1,906 SEM images were composed of approximately 44% Secondary Electron Imaging (SEI) or Low-Energy Electron Imaging (LEI) images and 56% Large-Angle Backscattered Electron Imaging (LABE) images. Preliminary results indicate higher accuracies when using LABE images compared to SEI/LEI images for both statistical and deep machine learning applications. This work introduces a groundbreaking methodology that revolutionizes particle classification by splitting SEM particle images into sub-images for analysis. These sub-images are then utilized in a majority-rules voting schema to classify the overall particle present in the whole image. The final deep learning framework, enhanced with Shannon Entropy techniques, empowers morphology imagery analysts to reduce workload, standardize processes, and deliver unprecedented forensic explainability setting a new benchmark in and offering new insights for critical nuclear forensic applications

AFIT Designator

AFIT-ENG-MS-25-M-012

Comments

An embargo was observed for this posting.

Approved for Public Release, Distribution Unlimited. PA case number 88ABW-2025-0218

Download

Share

COinS