斑点图案
去相关
数字图像相关
亚像素渲染
计算机科学
光学
人工智能
散斑成像
像素
图像处理
算法
计算机视觉
物理
图像(数学)
作者
Xiaoying Zhang,Yuxi Chi,Liting Yu,Qianqian Wang,Bing Pan
标识
DOI:10.1016/j.optlastec.2023.109913
摘要
Laser speckles have served as an alternative to artificial speckles for digital image correlation (DIC) measurement in some cases due to the advantages of non-invasive measurement, simplicity and high efficiency in preparation. Regrettably, in using laser speckle DIC, it unavoidably encounters the notorious decorrelation problem when the object surface undergoes a relatively large motion and deformation, which will degrade the accuracy of DIC matching or even destroy the measurement. To realize large deformation measurements, the approach of updating reference image (also termed as incremental DIC) is usually adopted to mitigate the decorrelation problem of laser speckle. However, once the reference image is updated frequently during the incremental DIC calculation, additional errors due to intensity interpolation at the subpixel positions in updated reference images will accumulate over the updating times. To address this issue, an improved calculation scheme composed of incremental DIC with the nearest integer pixel offset and Gaussian pre-filtering is proposed for laser speckle DIC measurement. This involves processing the speckle images with a Gaussian low-pass filter before correlation analysis, and shifting the reference subsets in updated reference images to the nearest integer pixel positions instead of interpolating at subpixel positions during DIC calculation. Simulation tests have confirmed that this scheme effectively mitigates accumulated interpolation errors. Several real tests, including in-plane translation, uniaxial tensile, and high-temperature tests, also demonstrated the effectiveness of this scheme in eliminating the decorrelation issue in laser speckle DIC.
科研通智能强力驱动
Strongly Powered by AbleSci AI