Premium
DVC‐based image subtraction to detect microcracking in lightweight concrete
Author(s) -
Chateau C.,
Nguyen T. T.,
Bornert M.,
Yvonnet J.
Publication year - 2018
Publication title -
strain
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.477
H-Index - 47
eISSN - 1475-1305
pISSN - 0039-2103
DOI - 10.1111/str.12276
Subject(s) - voxel , digital image correlation , affine transformation , interpolation (computer graphics) , artificial intelligence , compression (physics) , computer science , subtraction , materials science , transformation (genetics) , computer vision , filter (signal processing) , image (mathematics) , mathematics , optics , physics , geometry , arithmetic , composite material , biochemistry , chemistry , gene
This paper presents an image subtraction technique based on digital volume correlation to detect and extract the complex network of microcracks that progressively developed in a lightweight concrete sample submitted in situ to uniaxial compression and imaged by X‐ray computed tomography. From local digital volume correlation measurements, performed only on positions with sufficient image contrast, the mechanical transformation is estimated at all voxels within the whole sample using an adjusted interpolation procedure that computes an affine approximation of the local transformation. The deformed image (containing cracks) is thus transformed back to the same frame as the reference image (without cracks) to compute the difference between both images, taking into account possible brightness and contrast adjustments. The resulting subtracted image reveals the path of cracks, which is clearly visible without the underlying heterogeneous microstructure of the concrete. The detection accuracy is here estimated to one tenth of a voxel, allowing early‐age cracks to be detected while they would barely have been noticed on the X‐ray computed tomography images. Segmentation of the crack network is also made much easier. To overcome a low signal‐to‐noise ratio for the tiniest cracks, a Hessian‐based filter is used to extract the complex crack network. The cracks can be directly located in the microstructure segmented in the reference image and compared for all loading steps to characterise their initiation and propagation.