Correction of Scanning Electron Microscope Imaging Artifacts in a Novel Digital Image Correlation Framework
Correction of Scanning Electron Microscope Imaging Artifacts in a Novel Digital Image Correlation Framework
复制标题
DOI：
10.1007/s11340-018-00469-w
复制
发表时间：
2019-04-01
影响因子：
2.4
通讯作者：
Geers, M. G. D.
中科院分区：
文献类型：
--
作者：
Maraghechi, S.;Hoefnagels, J. P. M.;Geers, M. G. D.
数字图像相关 (DIC) 和扫描电子显微镜 (SEM) 的结合能够基于 SEM 的高空间分辨率和 DIC 的亚像素精度提取高分辨率全场位移数据。然而，SEM 图像可能会出现大量的成像伪影，这可能会严重影响从这些图像测量的位移和应变的准确性。当前的研究提出了一个统一的通用框架来纠正 SEM 伪影的三种主要类型，即空间畸变、漂移畸变和扫描线移位。伪影场与机械变形一起测量，以最大限度地减少后者中伪影引起的误差。为此，集成 DIC (IDIC) 扩展了一系列分层映射函数，这些函数描述了成像过程与力学的相互作用。推导了基于这些映射函数的新 IDIC 公式，并通过大量虚拟实验测试了该框架的潜力。研究了图像中噪声的影响以及伪影场的不同正则化选项。对于所有研究的情况，在噪声水平高达 5% 的情况下测量的机械位移场的误差在通常的 DIC 精度范围内，而如果忽略伪影，则误差超过 4 个像素。对三种不同放大倍数的真实 SEM 图像进行的验证证实，所有三个畸变场均被准确捕获。所有虚拟和真实实验的结果证明了所提出方法的准确性及其在收敛方面的鲁棒性。
The combination of digital image correlation (DIC) and scanning electron microscopy (SEM) enables to extract high resolution full field displacement data, based on the high spatial resolution of SEM and the sub-pixel accuracy of DIC. However, SEM images may exhibit a considerable amount of imaging artifacts, which may seriously compromise the accuracy of the displacements and strains measured from these images. The current study proposes a unified general framework to correct for the three dominant types of SEM artifacts, i.e. spatial distortion, drift distortion and scan line shifts. The artifact fields are measured alongside the mechanical deformations to minimize the artifact induced errors in the latter. To this purpose, Integrated DIC (IDIC) is extended with a series of hierarchical mapping functions that describe the interaction of the imaging process with the mechanics. A new IDIC formulation based on these mapping functions is derived and the potential of the framework is tested by a number of virtual experiments. The effect of noise in the images and different regularization options for the artifact fields are studied. The error in the mechanical displacement fields measured for noise levels up to 5% is within the usual DIC accuracy range for all the cases studied, while it is more than 4 pixels if artifacts are ignored. A validation on real SEM images at three different magnifications confirms that all three distortion fields are accurately captured. The results of all virtual and real experiments demonstrate the accuracy of the methodology proposed, as well as its robustness in terms of convergence.