A Review of Computer Vision-Based Structural Deformation Monitoring in Field Environments

Yizhou Zhuang, Weimin Chen, Tao Jin, Bin Chen, He Zhang, Wen ZhangSensors 2022 · DOI

TL;DR

This review paper provides a comprehensive overview of computer vision-based structural deformation monitoring techniques for field environments. It analyzes the critical factors affecting measurement accuracy, namely hardware/environmental impacts and target tracking algorithms, and discusses existing solutions and limitations for long-term, stable monitoring.

What Problem Does This Paper Address?

Structural health monitoring (SHM) is crucial for ensuring the safety and longevity of transportation infrastructure. While traditional sensor networks have limitations in terms of cost and installation complexity, computer vision offers a promising non-contact alternative. However, transitioning these techniques from controlled lab environments to real-world field applications presents significant challenges. These include the impact of environmental factors (temperature, lighting, vibration), hardware limitations (camera quality, lens), and the robustness of target tracking algorithms. This paper aims to consolidate the current state of research, identify these challenges, and propose solutions for reliable, long-term structural deformation monitoring using computer vision in field settings.

Methods

The paper categorizes computer vision-based monitoring systems into two main components: image acquisition and image processing.

Image Acquisition: This involves selecting appropriate cameras (CCD vs. CMOS, considering resolution, frame rate, and shutter type), lenses (fixed-focus vs. zoom, distortion, focal length), and targets (artificial vs. natural, their patterns, and robustness).

Image Processing: This encompasses several key stages:

  • Camera Calibration: Essential for converting pixel coordinates to real-world measurements. Methods discussed include the full projection matrix, planar homography matrix, and scale factor, each with varying complexity and applicability.
  • Feature Extraction and Target Tracking: Algorithms like shape matching, feature point matching (e.g., SIFT, SURF, BRIEF), optical flow, and Digital Image Correlation (DIC) template matching are reviewed for their strengths and weaknesses in identifying and following targets.
  • Deformation Calculation: Transforming pixel displacement into actual displacement, which is heavily influenced by the preceding calibration and tracking steps.

The paper also delves into the impact of hardware (camera, targets) and environmental factors (optical refraction, camera motion, illumination changes, occlusion) on measurement accuracy, and discusses various algorithms and techniques to mitigate these issues.

Datasets

Not discussed in the paper. The paper focuses on reviewing existing research and methodologies rather than presenting new experimental datasets.

Key Insights and Ideas

  • Field vs. Lab: The transition from controlled lab environments to dynamic field conditions is the primary hurdle for computer vision-based SHM. Environmental variability (temperature, light, vibration) significantly degrades performance.
  • Two Pillars of Accuracy: Measurement accuracy hinges on two critical aspects: the quality of image acquisition (hardware and environmental resilience) and the sophistication of the target tracking algorithm.
  • Target Choice Matters: Artificial targets generally offer higher accuracy and robustness, but their installation can be problematic. Natural targets are easier to use but are more susceptible to environmental changes.
  • Algorithm Trade-offs: Different tracking algorithms offer varying balances between speed, accuracy, and robustness to illumination changes and partial occlusion. There's no one-size-fits-all solution.
  • Sub-pixel is Key for Precision: Achieving high measurement accuracy often relies on sub-pixel estimation techniques, but these can be computationally intensive and still susceptible to real-world noise.
  • Long-Term Stability is the Holy Grail: While significant progress has been made, achieving truly long-term, stable, and reliable deformation monitoring in field environments remains an ongoing challenge.

Key Results

The paper synthesizes results from numerous studies, highlighting:

  • Accuracy: Various methods report errors ranging from sub-millimeter to a few millimeters at different distances. For instance, some studies achieve errors less than 0.01 mm at short distances, while others report average errors around 0.5674 mm at 300 m.
  • Efficiency: Improvements in algorithms have led to significant speedups. For example, some methods are reported to be 5 to 120 times faster than traditional approaches without sacrificing accuracy.
  • Robustness: Techniques like adaptive optical-turbulence error filters have shown to reduce measurement errors by about 67.5%. Background modification methods can reduce measurement errors from 44.1 mm to 1.1 mm.
  • Application Scenarios: The paper maps algorithms to specific applications, such as shape matching for cable structures and stadiums, and DIC for long-span bridges.

Limitations and Future Work

  • Long-Term Reliability: The paper acknowledges that while short-term measurements are well-established, the uncertainty and reliability of computer vision systems for long-term monitoring (months to years) require further research.
  • Data Acquisition and Transmission: The need for specialized personnel for installation, maintenance, and data handling presents a challenge for widespread adoption.
  • Time Synchronization: For multi-camera systems, precise time synchronization is crucial and requires further development.
  • Uncertainty Evaluation: Developing robust methods for evaluating the uncertainty of vision sensors in long-term applications is a key area for future work.
  • Simpler Programs/Devices: The authors suggest the need for simpler programs or devices to facilitate the long-term application of these technologies in practical engineering.

Summary updated

Abstract

Computer vision-based structural deformation monitoring techniques were studied in a large number of applications in the field of structural health monitoring (SHM). Numerous laboratory tests and short-term field applications contributed to the formation of the basic framework of computer vision deformation monitoring systems towards developing long-term stable monitoring in field environments. The major contribution of this paper was to analyze the influence mechanism of the measuring accuracy of computer vision deformation monitoring systems from two perspectives, the physical impact, and target tracking algorithm impact, and provide the existing solutions. Physical impact included the hardware impact and the environmental impact, while the target tracking algorithm impact included image preprocessing, measurement efficiency and accuracy. The applicability and limitations of computer vision monitoring algorithms were summarized.
BibTeX
@inproceedings{zhuang2022review,
  author = {Yizhou Zhuang and Weimin Chen and Tao Jin and Bin Chen and He Zhang and Wen Zhang},
  title = {{A Review of Computer Vision-Based Structural Deformation Monitoring in Field Environments}},
  booktitle = {Sensors},
  year = {2022},
  doi = {10.3390/s22103789},
}
Figure