Theses and Dissertations

ORCID

https://orcid.org/0009-0007-9044-747X

Advisor

Swan, J. Edward

Committee Member

Carruth, Daniel

Committee Member

Bethel, Cindy

Date of Degree

5-10-2024

Original embargo terms

Immediate Worldwide Access

Document Type

Graduate Thesis - Open Access

Major

Computer Science (Research Computer Science)

Degree Name

Master of Science (M.S.)

College

James Worth Bagley College of Engineering

Department

Department of Computer Science and Engineering

Abstract

This study addresses the challenge of lens obfuscations in off-road autonomous vehicles, which compromise the essential visual inputs for safe navigation. Using a tiered approach, the research employs neural network architectures for preliminary image classification, semantic segmentation, and image-to-image translation to rectify obscured visual inputs. Initial classification using MobileNetV2 sets the stage for U-Net-driven semantic segmentation to identify obfuscated regions, followed by a modified Pix-to-Pix model for image restoration. The evaluation showcases promising results in improving visual clarity, marking a significant stride towards enhancing autonomous vehicle operational robustness in off-road environments. This work lays a foundation for future explorations into advanced neural network architectures for real-time implementations in off-road terrains.

Share

COinS