Evaluation of Ballast Degradation Using Computer Vision and Comparison with Ground Penetrating Radar

Tuesday, June 10, 2025

3:15 PM – 3:30 PM MT

Primary Author(s)

Youngdae Kim

Ph.D. Student
University of Illinois Urbana-Champaign
Urbana, IL, United States

Co-Author(s)

Kelin Ding, MS

Ph.D. Student
University of Illinois Urbana-Champaign
Urbana, IL, United States
Yongsung Koh, PhD (he/him/his)

Postdoctoral Research Associate
University of Illinois Urbana-Champaign
Champaign, IL, United States
Taeyun Kong, MSc

Ph.D. Student
University of Illinois Urbana Champaign
Urbana, IL, United States
Erol Tutumluer, Ph.D.

Abel Bliss Professor in Engineering
University of Illinois Urbana-Champaign
Urbana, Illinois, United States

Presenter(s)

Erol Tutumluer, Ph.D.

Abel Bliss Professor in Engineering
University of Illinois Urbana-Champaign
Urbana, Illinois, United States

Co-Author(s)

Jeremy J. Beasley, n/a

ERDC Railroad Track Evaluation Program Team Leader
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi, United States
AC

Abby Cisko, Ph.D.

Research Mechanical Engineer
U.S. Army Engineer Research and Development Center
Vicksburg, Mississippi, United States
Michael J. Harrell, P.E.

Principal Engineer, Vice President
Applied Research Associates, Inc.
Champaign, Illinois, United States

Submitter(s)

Youngdae Kim

Ph.D. Student
University of Illinois Urbana-Champaign
Urbana, IL, United States

Abstract: Ballast aggregates provide support for rails and crossties, effectively distributing wheel loads and ensuring proper drainage within the ballasted track substructure. To maintain these functions, it is required to monitor the level of ballast degradation and assess track conditions. Traditional methods for evaluating ballast, such as visual inspections or collecting samples for sieve analysis, are often subjective and labor-intensive. Furthermore, it may face challenges in fully capturing the variability of ballast degradation along the track with traditional approaches. To overcome these limitations, a Ballast Scanning Vehicle (BSV) has been recently developed at the University of Illinois to automate ballast condition assessments using computer vision techniques. This study aims to evaluate ballast degradation using BSV technology and validate its performance by comparing it to Ground Penetrating Radar (GPR) assessments. The BSV testing was carried out on both clean and degraded ballast at the Ft. Campbell Army installation in Kentucky. A line scan camera was used to capture the plan and profile views of the ballast shoulder after trenching, performed using a ballast regulator, with samples collected at multiple depths. The captured RGB images were processed to estimate Fouling Index (FI) and Particle Size Distribution (PSD), using a computer vision algorithm. The results were compared against laboratory sieve analysis. Additionally, the FI values derived from BSV data were compared to those obtained from GPR analysis. This paper offers insights into the application of automated computer vision systems for more efficient and consistent ballast degradation assessment. Furthermore, the comparison between BSV and GPR results highlights the strengths and limitations of each technique, contributing to a more reliable and comprehensive approach to track condition assessment and railway infrastructure monitoring.

Learning Objectives:

Attendees can expect to learn the following from this session:

Upon completion, participants will be able to contrast traditional ballast evaluation methods with automated computer vision technology, highlighting the advantages and limitations of each approach for assessing ballast degradation.
Upon completion, participants will be able to explain how Ballast Scanning Vehicle technology utilizes computer vision to assess ballast condition and estimate Fouling Index and Particle Size Distribution.
Upon completion, participants will be able to compare the ballast condition assessment results obtained from BSV with those from GPR analysis, identifying the advantages and limitations of each method.