Building 3D Microstructure Database Using an Advanced Metallographic Serial Sectioning Technique and Robust 3D Segmentation Tools

Abstract

Building a 3D microstructure database is gaining rapid attention as it provides material specific key knowledgebase toward ICME approaches. Recent advances in automated metallographic serial sectioning technology such as RoboMet.3D, has significantly improved metallographic procedures and control systems for the acquisition of high quality metallographs in the form of serial sections. However, the image registration for 3D reconstruction and the segmentation of actual microstructural attributes from the raw data is still challenging, particularly for polycrystalline materials. The present study demonstrates systematic efforts to improve the automated metallographic serial sectioning technique for the 3D microstructure data. In particular, comprehensive image-processing tools were developed for the segmentation of complex 3D microstructures. Its capability of handling various complex 3D microstructure data (from polycrystals to composites) is also demonstrated. Geometric and gray-scale features are calculated using voxel values and their scale-space relational analysis. We propose to build a volumetric image database that allows qualitative and quantitative comparison of different materials and their structural properties measured using image processing algorithms and micrographs.