Fuel Cells and Solid State Chemistry Division, Risø National Laboratory for Sustainable Energy, Technical University of Denmark, Building 778, Frederiksborgvej 399, 4000 Roskilde, Denmark.
Ultramicroscopy. 2010 Feb;110(3):216-28. doi: 10.1016/j.ultramic.2009.11.013. Epub 2009 Nov 26.
Routine use of quantitative three dimensional analysis of material microstructure by in particular, focused ion beam (FIB) serial sectioning is generally restricted by the time consuming task of manually delineating structures within each image slice or the quality of manual and automatic segmentation schemes. We present here a framework for performing automatic segmentation of complex microstructures using a level set method. The technique is based on numerical approximations to partial differential equations to evolve a 3D surface to capture the phase boundaries. Vector fields derived from the experimentally acquired data are used as the driving forces. The framework performs the segmentation in 3D rather than on a slice by slice basis. It naturally supplies sub-voxel precision of segmented surfaces and allows constraints on the surface curvature to enforce a smooth surface in the segmentation. Two applications of the framework are illustrated using solid oxide cell materials as examples.
常规使用通过聚焦离子束（FIB）连续切片等方法对材料微观结构进行定量三维分析，通常受到手动描绘每个图像切片内结构的耗时任务或手动和自动分割方案的质量的限制。我们在这里提出了一种使用水平集方法进行复杂微观结构自动分割的框架。该技术基于偏微分方程的数值逼近方法，通过演化一个 3D 曲面来捕获相边界。从实验获得的数据中提取的向量场用作驱动力。该框架在 3D 中执行分割，而不是逐片进行。它自然提供了分段表面的亚像素精度，并允许对表面曲率施加约束，以在分割中保持表面光滑。使用固体氧化物电池材料作为示例，说明了该框架的两个应用。
