127 / 2018-08-24 23:28:27

Reading Local Structure for Ferroelectric Ceramic

image segmentation,Hough transform,ferroelectric materials,CBED pattern

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. Here, a computational method is proposed to segment patterns and analyze 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 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.