Vision and Vibration Data Fusion-Based Structural Dynamic Displacement Measurement with Test Validation.

Area of Science:

• Structural Engineering
• Computer Vision
• Sensor Fusion

Background:

• Dynamic measurement of structural deformation is crucial for structural health monitoring.
• Traditional contact-type displacement monitoring has limitations in practical applications due to setup requirements.
• Computer vision offers non-contact, cost-effective displacement monitoring but is sensitive to environmental factors.

Purpose of the Study:

• To develop a data fusion method combining contact acceleration monitoring and non-contact vision-based displacement recognition.
• To improve the accuracy and sampling rate of dynamic displacement measurements for structural health monitoring.
• To overcome the limitations of existing vision-based methods influenced by lighting and resolution.

Main Methods:

• Utilized high dynamic sampling rate of contact acceleration sensors.
• Implemented an improved Kanade-Lucas-Tomasi (KLT) feature tracker algorithm.
• Employed asynchronous multi-rate Kalman filtering for data fusion and displacement estimation.

Main Results:

• The proposed method significantly improves the displacement sampling rate compared to vision-only techniques.
• High-frequency vibration information can be collected more effectively.
• Achieved a normalized root mean square error of less than 2% for displacement estimation.

Conclusions:

• The data fusion approach enhances dynamic displacement monitoring capabilities for structural health.
• This method provides a more robust and accurate solution for structural deformation identification.
• The integration of acceleration and vision data offers superior performance in dynamic displacement measurement.