基于边缘检测与数字图像相关的桥梁结构变形测试

Structural Deformation Test Based on Edge Detection and Digital Image Correlation

数字图像处理(DIP)技术在机械制造和自动化工程等领域用于工厂化精确尺寸测量,其特点包括非接触、高速、高精度、低成本。而数字图像相关(DIC)技术可实现工程结构的亚像素级非接触实时形变测试。研究验证了DIP算法在工程结构位移测量的可行性。引入Canny和Zernike矩两种边缘识别算法,通过对粘贴于测试点的特定图案靶标进行边缘检测,对检测结果进行几何学拟合实现亚像素定位,对比不同时刻的靶标位置,从而获得结构位移时间变化。通过模拟试验、实验室试验和实桥试验,比较了基于DIP形变测试算法和DIC形变测试算法的精度、速度和误差。研究表明,基于DIP的结构形变测试在模拟试验中精度可达0.1 pixel,实验室条件下可达0.1 mm。Canny和Zernike矩算法的精度相近,低于DIC算法;DIC算法子集扩大时计算速度显著降低。Canny和DIC算法速度明显快于Zernike矩算法,通过CUDA并行加速,Zernike矩算法速度与小子集的DIC算法和Canny算法相当。在实桥试验中,DIC算法测试误差小于0.05 mm,精度最高;Canny和Zernike矩算法精度可达亚毫米级,适用于某些工程结构位移测试场景。

Digital Image Processing(DIP)is widely utilized in mechanical manufacturing and automation engineering for precise large-scale dimensional measurement.DIP′s non-contact,high-speed,high-precision,and cost-effective features make it a preferred choice.Concurrently,Digital Image Correlation(DIC)excels in real-time,sub-pixel-level deformation analysis in engineering structures.This study validates the feasibility of using DIP algorithms for quantifying structural displacements.We introduce two edge detection algorithms,Canny and Zernike moments,in this study.These algorithms identify edges onpatterned targets at measurement locations.Geometric fitting and sub-pixel localization are then applied to the results.Variations in target positions over time are scrutinized to assess structural displacement.We comprehensively compare the effectiveness of DIP-based and DIC-based deformation testing algorithms,considering precision,processing speed,and error rates.Our research reveals that DIP-based methods for assessing structural deformations exhibit a high level of precision,achieving an accuracy of 0.1 pixel in simulations and 0.1 millimeters in controlled laboratory conditions.Canny and Zernike moment algorithms offer comparable accuracy but fall short of DIC's precision.DIC's computational efficiency decreases with larger subsets,while Canny and DIC algorithms outperform Zernike moments in processing speed.By utilizing CUDA parallel processing,Zernike moment algorithms significantly enhance their speed,making them competitive with DIC and Canny algorithms,particularly in scenarios with smaller subsets.In practical bridge experiments,DIC achieves testing errors of less than 0.05 millimeters with the highest precision,while Canny and Zernike moment algorithms provide sub-millimeter precision,suitable for specialized engineering applications requiring structural displacement assessments.