The Alignment and Fusion of Multimodal 3D Serial Sectioning Datasets

As an example of data fusion in the context of 3D characterization of materials, this article demonstrates the procedures necessary to align and fuse separate imaging modes, traditional backscattered electron imaging (BSE) and electron backscattered diffraction mapping (EBSD), from serial-sectioning data. The fused data form a unified 3D reconstruction of additively manufactured 316L stainless steel processed by laser powder-bed fusion. We show that, by combining the relatively low-information yet high-fidelity BSE image stack with the more data-rich yet spatially distorted EBSD maps, the 3D reconstruction can leverage the strengths of both imaging techniques. The fully automated alignment procedures and frameworks rely on a number of optimized image warping techniques, with the result that spatial alignment errors are on the order of 0–3 μm\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu {\hbox {m}}$$\end{document} within a region of interest that is >1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$> 1$$\end{document} mm.