Global optimization for accurate determination of EBSD pattern centers

TL;DR

This paper introduces a robust global optimization method combining a global search algorithm and multi-pattern averaging to accurately determine EBSD pattern centers, overcoming noise and binning challenges.

Contribution

It presents a novel approach for precise pattern center determination in EBSD by addressing the noisy and sloppy parameter landscape with a combined global search and averaging technique.

Findings

Accurately determines pattern centers in simulated noisy and binned patterns.

Effectively detects pattern center changes in experimental noisy datasets.

Source code is publicly available for implementation.

Abstract

Accurate pattern center determination has long been a challenge for the electron backscatter diffraction (EBSD) community and is becoming critically accuracy-limiting for more recent advanced EBSD techniques. Here, we study the parameter landscape over which a pattern center must be fitted in quantitative detail and reveal that it is both sloppy and noisy, which limits the accuracy to which pattern centers can be determined. To locate the global optimum in this challenging landscape, we propose a combination of two approaches: the use of a global search algorithm and averaging the results from multiple patterns. We demonstrate the ability to accurately determine pattern centers of simulated patterns, inclusive of effects of binning and noise on the error of the fitted pattern center. We also demonstrate the ability of this method to accurately detect changes in pattern center in an experimental dataset with noisy and highly binned patterns. Source code for our pattern center fitting algorithm is available online.