A Dictionary Approach to Electron Backscatter Diffraction Indexing.

Area of Science:

• Materials Science
• Crystallography
• Data Analysis

Background:

• Electron Backscatter Diffraction (EBSD) is crucial for characterizing polycrystalline materials.
• Accurate indexing of grain and subgrain structures is essential for understanding material properties.
• Current methods face challenges in efficient and unified segmentation, anomaly detection, and orientation indexing.

Purpose of the Study:

• To propose a novel framework for indexing grain and subgrain structures in EBSD patterns.
• To develop a unified approach for segmentation, anomaly detection, and crystal orientation indexing.
• To quantify uncertainty in the analysis of EBSD data.

Main Methods:

• A dictionary of patterns is generated by discretizing a dynamical forward model onto a dense grid of orientations.
• Measured EBSD patterns are matched against the dictionary to identify similar patterns.
• An unsupervised binary decision tree (DT) classifier, utilizing the statistical distribution of closest matches, is employed for segmentation and anomaly detection.
• Orientation indexing is achieved by computing the mean orientation of closest matches using a maximum likelihood approach with Von Mises-Fisher distributions.

Main Results:

• The framework successfully identifies grain boundaries and anomalous regions within EBSD patterns.
• It enables accurate indexing of crystal orientations by estimating the mean orientation of closest matches.
• The dictionary matching approach provides a unified method for segmentation, anomaly detection, and indexing.

Conclusions:

• The proposed dictionary matching framework offers a robust and unified approach for analyzing EBSD data.
• It enhances the accuracy of grain structure indexing and anomaly detection in polycrystalline materials.
• The method inherently provides uncertainty quantification, adding value to the analysis.