Why Vision-based Structural Health Monitoring?

Transportation infrastructures, apart from experiencing continuous degradation of material performance, are also subjected to human and natural loads (e.g., repeated loads from vehicles, wind, and earthquakes). These loads can result in cumulative or sudden damage, threatening safe operation and potentially leading to disasters. Traditional structural health monitoring (SHM) methods often rely on contact-based sensors, which can be costly, time-consuming, and limited in coverage. Vision-based monitoring, powered by computer vision and AI, offers a non-contact, cost-effective, and scalable solution to assess structural conditions, enabling early detection of damage and ensuring the safety and longevity of critical infrastructure.

Objectives

The COSMO project aims to:

• Inspect and diagnose infrastructure surface defects (e.g., corrosion, cracks, loose and missing parts).
• Measure structural dynamics responses (e.g., displacements, vibrations) and estimate key properties (e.g., natural frequencies, damping ratio) to assess structural integrity.
• Develop accurate, robust, and automated monitoring systems for transportation infrastructures to enhance safety and reduce maintenance costs.

Research Achievements

The COSMO project has achieved several milestones, including:

• Bridge Rivet Inspection and Diagnosis: Automated detection and assessment of rusted, loose or missing rivets in bridge structures. [Articles: CACAIE, EngStruct]
• Corrosion Detection and Diagnosis: Identification and quantification of corrosion in metal components using transformer-based segmentation. [Article: CSHM]
• Camera Motion Correction for Optical Measurements: A novel 6-DOF camera motion correction method to enhance the accuracy and robustness of optical measurements. [Article: MSSP]
• Railway Catenary Uplift Measurements and Damping Estimation: Visual monitoring of in-service railway catenary systems to ensure operational safety. [Articles: MSSP, MMT, JSV]

Future Work

• Develop multi-sensor fusion techniques to combine vision-based data with other SHM methods for enhanced reliability.
• Integrate AI-driven predictive analytics to forecast structural degradation and optimize maintenance schedules.
• Improve real-time processing capabilities for large-scale monitoring of infrastructure networks.
• Expand the application of COSMO to other infrastructure types.

Collaboration

The project was a collaboration between NTNU, Norwegian Railway Directorate (Jernbanedirektoratet), and Bane NOR