Abstract
The irregular arch bridge is increasingly favored by designers for its aesthetic appeal. However, traditional displacement measurement methods fail to meet the high-precision, high-frequency requirements for health monitoring of such irregular arch bridge structures. Therefore, based on the Echeng Bridge in Huizhou City, a study proposes a close-range photogrammetry system using ellipse fitting and inverse perspective transformation principles. This system effectively eliminates displacement errors caused by shooting angles and improves measurement accuracy. Experimental results demonstrate that compared to traditional total station measurement methods, both methods yield highly consistent results with measurement accuracy reaching millimeter-level requirements. The study analyzes the feasibility and measurement accuracy of applying this visual measurement system to the Echeng Bridge, providing a theoretical basis for similar irregular arch bridges.
Introduction
Irregular arch bridges represent a pinnacle of modern infrastructure aesthetics, yet their complex geometries present significant challenges for structural health monitoring. These bridges experience intricate displacements and deformations under varying loads that traditional methods, such as leveling or total station measurements, struggle to capture efficiently. Traditional techniques are often labor-intensive, expensive, and lack the frequency required for real-time monitoring. While Digital Image Correlation (DIC) has emerged as a non-contact alternative, it remains sensitive to environmental noise and perspective distortion in outdoor engineering settings. This paper addresses these gaps by proposing a robust close-range visual measurement system.
Methodology
The Visual Measurement Pipeline
The proposed system follows a structured pipeline designed to translate raw visual data into precise structural displacement metrics. The process begins with image acquisition and moves through feature recognition and tracking before converting pixel-based movements into physical displacement values.
Feature Recognition and Subpixel Localization
To handle the complexities of an outdoor environment, the system employs a multi-stage recognition process:
- Image Preprocessing: Shi-Tomasi corner detection is used to automatically identify Regions of Interest (ROIs). To mitigate environmental noise, the system applies median and Gaussian filtering, followed by a Laplacian transformation to enhance target contrast.
- Ellipse Fitting: The system tracks elliptical markers rather than simple points. It utilizes an edge-following method to identify arc segments, which are then grouped using Hough transform principles. A "completeness score" is assigned to each candidate ellipse; only those with high scores undergo secondary fitting, ensuring reliability even under partial occlusion.
- Subpixel Localization: To surpass the physical resolution of the camera sensor, bilinear sub-pixel interpolation is used. This allows the system to mathematically estimate edge positions between physical pixels, significantly boosting accuracy.
Case Study: Echeng Bridge
The system's feasibility was tested on the Echeng Bridge in Huizhou City, China—a 460-meter, four-span continuous irregular inclined cable-stayed arch bridge. Researchers used a finite element model (midas Civil) alongside real-world field data to validate the system. A key insight from this study was the necessity of accounting for the camera's own movement; structural torsion can cause the camera's mounting point to shift by nearly 2 mm, a factor that must be compensated for in high-precision monitoring.
Key Results and Discussion
- Precision and Repeatability: At a distance of 100 meters, the system achieved a theoretical pixel accuracy of 0.35 mm. In practical field conditions, the measured positioning accuracy was 1.82 mm with a repeatability of 0.91 mm.
- Comparison with Traditional Methods: The theoretical accuracy of the visual system (0.35 mm) outperformed the angle measurement accuracy of a standard total station (0.49 mm) at the same distance.
- Engineering Viability: During the construction phase, the maximum expected arch displacement was 137.4 mm. The visual system's accuracy falls well within the 5% error margin (6.68 mm) required for safe engineering monitoring.
Conclusion
This study demonstrates that close-range photogrammetry is a highly feasible and accurate solution for monitoring irregular arch bridges. By utilizing sub-pixel interpolation and robust ellipse fitting, the system provides a cost-effective, high-frequency alternative to traditional surveying tools. While accuracy naturally degrades with distance, the system's performance on the Echeng Bridge provides a strong theoretical and practical foundation for its application in similar complex infrastructure projects.
Summary updated