Automatic Mesh Size Estimation in DVC for Images of Isotropic Materials
Zaira Manigrasso, Jan Aelterman, Wilfried R. Philips · Ghent University Academic Bibliography (Ghent University) · 2020
When non-rigid digital image or volume correlation (DIC/DVC) is performed, it is critical to correctly set a parameter that controls the control point spacing for the grid on which deformation is defined. In this paper we present a method to automatically estimate the best performing grid spacing parameter for DIC/DVC registration. The operating principle is that the optimal grid spacing parameter is a function of the image content: it may be estimated through determining the dominant feature/object size. In order to extract the information about the object size, the image volume has been first segmented, then the disconnected objects inside the images have been detected (using labeling technique) and lastly a classification of the objects has been made based on the number of the voxels of each object. The materials for a validation study arise from the practical necessity resulting of finding the best performing parameter for registration in materials sciences research. We show how an erroneous setting of the density of the control points leads to inaccurate registration. Furthermore, we demonstrate how the parameter predicted by our algorithm is indeed optimal, both in a quantitative sense, using Normalized Cross Correlation (NCC) as a measure, as well as qualitatively.