基于图像序列间相似性的晶粒组织图像分割方法  

Segmentation method of grain microstructure based on similarity among image sequence

作　　者：薛维华[1,2] 王浩[1] 刘国权[1,3] 冯敏楠 齐琳[1] 陈展鹏[1] 

机构地区：[1]北京科技大学材料科学与工程学院,北京100083 [2]辽宁工程技术大学材料科学与工程学院,阜新123099 [3]北京科技大学钢铁共性技术协同创新中心,北京100083

出　　处：《中国体视学与图像分析》2015年第4期400-407,共8页Chinese Journal of Stereology and Image Analysis

基　　金：国家自然科学基金(51371030)

摘　　要：针对采用系列截面法进行晶粒组织三维重建过程中的图像分割问题,提出了基于图像序列间相似性的三维分割方法并实际应用于纯铁晶粒组织的重建。采用边缘检测方法进行了晶粒组织截面晶界线的提取。基于截面图像序列的相似特征,提出了实现截面图像二维分割的晶粒边界闭合方法和实现晶粒组织三维分割的识别与跟踪相邻截面中晶粒目标的方法,完成了纯铁晶粒组织的三维重建。同时分析了上述分割方法的优缺点,讨论了应加以改进的方向。Aiming at the segmentation difficulty in the three-dimensional reconstruction of grain structure by means of serial sectioning, a set of segmentation methods based on the similarity among image sequences were put forward and used in the reconstruction of pure iron grain structure. Using edge detection method, the grain boundary lines in the two dimensional section of grain structure were extracted. Based on the simi- larity among the section image sequences, the grain boundary closing method which could implement two- dimensional segmentation on section image and the grain identifying and tracking method which could im- plement three-dimensional segmentation on grain microstructure are put forward. Using these methods, a three-dimensional pure iron grain structure has been reconstructed. Meanwhile, the merits and drawbacks of the above methods are analyzed, and their improvement directions are discussed

关 键 词：晶粒组织 图像分割 图像序列 相似性 边界闭合 目标跟踪 三维重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]