Applying deep learning in automatic and rapid measurement of lattice spacings in HRTEM images  

深度学习用于HRTEM图像晶格间距的快速自动测量

作　　者：Xiaoyang Zhu Zuoheng Zhang Yu Mao Yan Li Xiao Huang Ning Gu 祝晓阳;张作恒;毛宇;李艳;黄晓;顾宁(State Key Laboratory of Bioelectronics,Jiangsu Key Laboratory for Biomaterials and Devices,School of Biological Science and Medical Engineering,Southeast University,Nanjing 210009,China;Institute of Advanced Materials,Nanjing Tech University,Nanjing 211816,China)

机构地区：[1]State Key Laboratory of Bioelectronics,Jiangsu Key Laboratory for Biomaterials and Devices,School of Biological Science and Medical Engineering,Southeast University,Nanjing 210009,China [2]Institute of Advanced Materials,Nanjing Tech University,Nanjing 211816,China

出　　处：《Science China Materials》2020年第11期2365-2370,共6页中国科学（材料科学（英文版）

基　　金：supported by the National Key Research and Development Program of China(2017YFA0104302);the National Natural Science Foundation of China(61821002,51832001 and 31800843)。

摘　　要：Deep learning has been playing increasingly important roles in image processing analysis.The advancement of materials science and engineering has driven the rapid development of transmission electron microscopy(TEM)and related techniques.Efficiently and accuratelyi ana-lyzing TEM data,which are usually in the form of image sequences,has become increasingly important.在图像处理与分析领域,深度学习发挥着日渐重要的作用.我们将深度学习应用于高分辨率透射电子显微镜(HRTEM)图像晶面间距的自动快速测量中,通过对随机样本数据训练得到具有U型结构的神经网络,开发了一种新的图像处理方法.本方法能够自动提取快速傅里叶变换(FFT)图像中的衍射斑点,进一步在计算机视觉技术的协助下,可以自动计算与FFT图像中被识别的衍射点相对应的晶格间距,并与标准晶体结构数据进行比较.以Fe3O4纳米粒子的HRTEM图像为例,用本方法进行自动测量的晶格间距与手动测量的晶格间距相比,误差小于1%.我们的工作证明了深度学习技术在协助晶体材料发展方面的巨大潜力.

关 键 词：自动测量 计算机视觉技术 图像处理方法 深度学习 FE3O4纳米粒子 随机样本 晶体材料 晶面间距 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]