Non-Target Structural Displacement Measurement Using Reference Frame-Based Deepflow.

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Structural health monitoring (SHM) requires accurate displacement measurement, which is challenging in field conditions.
• Existing methods for dynamic displacement measurement are often costly, labor-intensive, and lack precision for small movements.
• Computer vision (CV)-based methods offer a cost-effective, remote, and accurate alternative, but non-target-based approaches face challenges like feature point issues and drift.

Purpose of the Study:

• To develop an improved non-target-based CV method for structural displacement measurement.
• To address limitations of existing CV methods, including insufficient feature points, detection errors, occlusion, and drift.
• To enable accurate displacement tracking in regions lacking distinct natural features.

Main Methods:

• A reference frame-based Deepflow algorithm was integrated with masking and signal filtering.
• The method allows users to select points of interest (POIs) for displacement tracking, even in low-gradient image areas.
• Displacement is calculated directly, mitigating drift from accumulated measurement errors.

Main Results:

• The proposed method was experimentally validated on a cantilevered beam under ambient and occluded conditions.
• Accuracy was compared against a reference laser displacement sensor, demonstrating high precision.
• The technique successfully extracted structural displacement in regions without distinct natural features.

Conclusions:

• The developed reference frame-based CV algorithm provides a drift-free and accurate solution for non-target-based displacement measurement.
• This method enhances flexibility in SHM, allowing displacement monitoring on diverse structural surfaces.
• The approach offers a cost-effective and efficient alternative for measuring small dynamic displacements in real-world scenarios.