Pan Bing, Wang Zhaoyang, Lu Zixing
Institute of Solid Mechanics, Beijing University of Aeronautics & Astronautics, Beijing, China.
Opt Express. 2010 Jan 18;18(2):1011-23. doi: 10.1364/OE.18.001011.
Digital image correlation (DIC) is an easy-to-implement yet powerful optical metrology for deformation measurement. The technique measures the displacement of a point of interest by matching the subsets surrounding the same point located in the reference image and the deformed image. Although the technique is simple in principle, the existing DIC technique has several deficiencies. For example, for the points located near or at the boundaries of a specified region of interest (ROI), the selected square subsets surrounding these points may contain unwanted or foreign pixels from background image or other regions. In the existing DIC method, these points are either intentionally excluded from calculation or automatically removed after calculation, and leads to the absence of deformation information for the boundary points. Besides, existing DIC technique is prone to yield erroneous measurement for specimen with geometric discontinuities. In this paper, two approaches are developed to overcome the deficiencies of existing DIC technique. First, a modified Zero-mean Normalized Sum of Squared Differences (ZNSSD) criterion is defined for the correlation analysis of subsets surrounding the boundary points. Second, considering the possible complex shape of the ROI, a scanning strategy guided by the correlation coefficients of computed points is proposed to ensure reliable computation between consecutive points. With these two measures, the deformation of all the points including those located near or at the ROI boundaries can be automatically, reliably, and accurately determined. The improved DIC technique is universally applicable to the genuine full-field deformation measurement of objects with complex or arbitrary shapes. Two typical experimental image pairs are processed to evaluate the performance of the proposed method, and the results successfully demonstrate its effectiveness and practicality.
数字图像相关（DIC）是一种易于实现且功能强大的用于变形测量的光学计量方法。该技术通过匹配参考图像和变形图像中位于同一点周围的子集来测量感兴趣点的位移。尽管该技术原理简单，但现有的DIC技术存在一些缺陷。例如，对于位于指定感兴趣区域（ROI）边界附近或边界处的点，围绕这些点选择的方形子集可能包含来自背景图像或其他区域的不需要的或外来的像素。在现有的DIC方法中，这些点要么在计算中被有意排除，要么在计算后被自动去除，这导致边界点的变形信息缺失。此外，现有的DIC技术对于具有几何不连续性的试样容易产生错误的测量结果。本文提出了两种方法来克服现有DIC技术的缺陷。首先，为边界点周围子集的相关性分析定义了一种改进的零均值归一化平方差和（ZNSSD）准则。其次，考虑到ROI可能的复杂形状，提出了一种由计算点的相关系数引导的扫描策略，以确保连续点之间的可靠计算。通过这两种措施，可以自动、可靠且准确地确定所有点的变形，包括那些位于ROI边界附近或边界处的点。改进后的DIC技术普遍适用于具有复杂或任意形状物体的真正全场变形测量。处理了两对典型的实验图像来评估所提方法的性能，结果成功证明了其有效性和实用性。
