School of Geosciences and Info-Physics, Central South University, Changsha 410083, China.
Sensors (Basel). 2023 Feb 8;23(4):1884. doi: 10.3390/s23041884.
Computer vision-based displacement measurement techniques are increasingly used for structural health monitoring. However, the vision sensors employed are easily affected by optical turbulence when capturing images of the structure, resulting in displacement measurement errors that significantly reduce the accuracy required in engineering applications. Hence, this paper develops a multi-measurement point method to address this problem by mitigating optical-turbulence errors with spatial randomness. Then, the effectiveness of the proposed method in mitigating optical-turbulence errors is verified by static target experiments, in which the RMSE correction rate can reach up to 82%. Meanwhile, the effects of target size and the number of measurement points on the proposed method are evaluated, and the optimal target design criteria are proposed to improve our method's performance in mitigating optical-turbulence errors under different measurement conditions. Additionally, extensive dynamic target experiments reveal that the proposed method achieves an RMSE correction rate of 69% after mitigating the optical-turbulence error. The experimental results demonstrate that the proposed method improves the visual displacement measurement accuracy and retains the detailed information of the displacement measurement results.
基于计算机视觉的位移测量技术在结构健康监测中得到了越来越广泛的应用。然而，在对结构进行图像采集时，所使用的视觉传感器很容易受到光湍流的影响，导致位移测量误差，从而显著降低工程应用中所需的精度。因此，本文提出了一种多测量点方法，通过减轻空间随机性的光湍流误差来解决这个问题。然后，通过静态目标实验验证了所提出方法减轻光湍流误差的有效性，其中 RMSE 校正率可达到 82%。同时，评估了目标大小和测量点数量对所提出方法的影响，并提出了最优目标设计准则，以提高我们的方法在不同测量条件下减轻光湍流误差的性能。此外，广泛的动态目标实验表明，所提出的方法在减轻光湍流误差后可实现 69%的 RMSE 校正率。实验结果表明，所提出的方法提高了视觉位移测量精度，并保留了位移测量结果的详细信息。
