基于变分水平集方法获取岩石的细观结构

Obtaining rock's mesostructure based on variational level set method

岩土工程材料的细观结构特征会在很大程度上影响其宏观力学行为。随着数字图像处理技术的逐渐成熟,基于变分水平集方法的偏微分方程(PDE)方法在图像处理中越来越受重视。在岩石数字图像的基础上,运用基于变分水平集方法的C-V模型,结合图像的背景填充技术实现花岗岩图片的3相分割(长石、石英和云母),为进一步力学分析提供条件。上述方法可以量化岩石细观结构,最终获得能够准确反映岩石内部特征的图片,为将来的有限元分析提供真实几何模型。处理结果表明,通过选取适当的迭代次数和数字图像处理技术,该法能够获得较高质量的岩石细观结构量化图。

The geotechnical material＇s actual mesostructure is significant in that its characteristics will largely affect its macroscopic mechanical behavior. With the digital image processing technology being more and more mature, the partial differential equations （PDE） method based on variation level set attracts increasing attention meanwhile finds increasingly more application in the field of image processing. C-V model, as a frequently used method, is a typical one among those using the PDE method. This paper deals with how to use C-V model combined with background filling technology to realize the three-phase split （feldspar, quartz and mica） of the granite pictures after the digital image of the rock has been got; thus to provide the conditions for further mechanical analysis. The purpose of quantifying the mesostructure of rock can be achieved by the method described in this paper, which ultimately can be used to get pictures that accurately reflect the internal characteristics of the rock, thus to provide real geometry model for finite element analysis in the future. The numerical value results indicate that the PDE method is very effective indeed. By selecting the appropriate number of iterations, it can get a higher quality of the rock mesostructure quantitative geometric images consistent with the real mesostructure images.