Reading Local Structure for Ferroelectric Ceramic by Convergent Beam Electron Diffraction and Artificial Intelligence Method

Z. X. He,W. B. Yan,J. Z. Xu,Y. Wang,T. X. Zhao,J. H. Gao
DOI: https://doi.org/10.1109/ceidp47102.2019.9009960
2019-01-01
Abstract:Crystal structures are always considered as origin of properties of dielectric materials, which poses importance for microscopic observation, especially for materials with peculiar local structures. Convergent beam electron diffraction (CBED) is a widely used informative method of reading crystal structure. However, identifying CBED images by traditional artificial methods is a time-consuming process accompanied with certain ambiguity, which limits its application. Here, we identify CBED images by artificial intelligence methods to solve that problem. The method includes segmenting patterns and analyzing mirror symmetry of CBED images. Canny edge detection and Hough transform are used to divide CBED patterns into discs individually, which can successfully extract common key information of all CBED images. Then the direct grey level symmetry similarity evaluation (DGLS) and texture symmetric similarity detection method based on principal component analysis (TSPCA) are both calculated to estimate mirror symmetry performance quantitatively. The average matching rate of accuracy between computer calculation and visual inspection is 96.0% for DGLS and 97% for TSPCA. This method provides a uniform standard for mirror symmetry identification and is more conducive and objective to the analysis of a large number of experimental CBED images, which is of great significance to promote the further research of reading local structure for ferroelectric ceramic.
What problem does this paper attempt to address?