Convolutional neural network-based method for real-time orientation indexing of measured electron backscatter diffraction patterns
Convolutional neural network-based method for real-time orientation indexing of measured electron backscatter diffraction patterns
复制标题
DOI：
10.1016/j.actamat.2019.03.026
复制
发表时间：
2019-05
期刊：
影响因子：
9.4
通讯作者：
Yu-Feng Shen;R. Pokharel;T. Nizolek;Anilesh Kumar;T. Lookman
中科院分区：
文献类型：
--
作者：
Yu-Feng Shen;R. Pokharel;T. Nizolek;Anilesh Kumar;T. Lookman
电子背散射衍射（EBSD）是从多晶材料中获得空间分辨微结构信息的最常用技术。我们已经开发了两种基于域变换和迁移学习的卷积神经网络方法来从电子背散射衍射图案重建晶体取向。我们的模型是强大的实验测量的图像噪声和索引方向一样快的最高EBSD扫描速率。我们证明了四元数范数度量是一个强有力的指标，用于评估的可靠性重建的情况下，地面真相。我们证明了当前的方法对钽样品的适用性。
Electron backscatter diffraction (EBSD) is the most commonly used technique for obtaining spatially resolved microstructural information from polycrystalline materials. We have developed two convolutional neural network approaches based on domain transform and transfer learning to reconstruct crystal orientations from electron backscatter diffraction patterns. Our models are robust to experimentally measured image noise and index orientations as fast as the highest EBSD scanning rates. We demonstrate that the quaternion norm metric is a strong indicator for assessing the reliability of the reconstructions in the absence of the ground truth. We demonstrate the applicability of the current methods on a tantalum sample.