Date of Award:

8-2026

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Civil and Environmental Engineering

Committee Chair(s)

Mohsen Zaker Esteghamati

Committee

Mohsen Zaker Esteghamati

Committee

Sierra Young

Committee

Matt Hebdon

Abstract

Traffic signs help drivers and pedestrians to understand roadway rules, warnings, and directions, which determine how road users should behave. When signs fade or delaminate, they become harder to read and interpret, which creates increased safety hazards. Transportation agencies usually deploy engineers to the field to inspect signs because they need to evaluate and document the sign’s condition based on established standards and engineering judgment. This process requires substantial time and cost and becomes challenging to apply across large highway networks.

This study evaluated two computer vision (CV) models for automated traffic sign condition assessment using highway images collected across Utah. The work focused on two common sign defects: (i) fading and (ii) delamination. The study compared the performance of these models based on accuracy-related measures and processing speed.

The results showed that CV can help identify traffic sign defects, especially when the defects appear clearly in the image, and the training data contains enough representative examples. However, the results also showed that the model performance depends on data quality, label consistency, and the number of defect examples available for training. These findings suggest that CV can support future traffic sign monitoring and provide a practical benchmark for automated traffic sign condition assessment. The study also highlights the need for stronger datasets and more consistent annotations to support reliable, scalable, and data-driven infrastructure monitoring.

Download

Share

COinS

Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .