Logo image
Back
A vision-assisted bridge weigh-in-motion approach for highway bridges using wireless strain sensing
Thesis   Open access

A vision-assisted bridge weigh-in-motion approach for highway bridges using wireless strain sensing

Martin Simbarashe Chiworeke
Master of Science (M.S.), Drexel University
15 Jun 2026
DOI:
https://doi.org/10.17918/00011434
pdf
Chiworeke_Martin_20262.49 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Bridge weigh-in-motion Influence lines Sensing Vehicle load estimation You Only Look Once Computer Vision
Bridge weigh-in-motion (B-WIM) is a structural-response-based approach for estimating vehicle axle loads and gross vehicle weight (GVW) from the measured response of in-service bridges. Although the basic concept of B-WIM is well established, its reliable application under field traffic conditions remains difficult. This is because accurate load identification depends strongly on vehicle kinematic inputs such as lane position, speed, and axle spacing. In practice, these quantities are not always available from structural sensing alone. This thesis addresses this limitation by developing a vision-assisted B-WIM framework that integrates roadside video, synchronized wireless strain measurements, lane-specific influence lines, and inverse load identification for estimating grouped axle loads and GVW of highway trucks crossing an in-service bridge. The proposed methodology begins with the extraction of vehicle kinematics from roadside video using You Only Look Once (YOLO) based vehicle and axle detection, followed by lane assignment, speed estimation, and axle-spacing estimation. Wirelessly collected and synchronized bridge strain measurements are baseline-corrected and filtered to isolate the quasi-static response and then used for load identification. A calibrated finite element model of the monitored bridge span is used to generate lane-specific strain influence lines at the instrumented girder locations. These inputs are combined within a classical influence-line-based B-WIM formulation using grouped axle loads and a non-negative least-squares solution to estimate event-level vehicle loading from the measured bridge response. A numerical sensitivity study is also performed to quantify the effects of uncertainty in speed and axle spacing on grouped axle-load and GVW estimates. The results show that the proposed framework can extract event-specific vehicle kinematics from roadside video, produce grouped axle-load and GVW estimates, and reproduce measured bridge strain responses with reasonable agreement under field conditions. The sensitivity study shows that uncertainty in speed estimation has the strongest effect on GVW estimation, while uncertainty in axle spacing estimation has a smaller effect on the total estimated weight but can significantly alter the grouped load distribution. These findings demonstrate the feasibility of integrating roadside computer vision, wireless strain sensing, and lane-specific influence lines into a practical vision-assisted B-WIM framework for field traffic load estimation and bridge monitoring.

Metrics

1 Record Views

Details

Logo image