基于形态学分水岭算法的骨料分割及建模

Segmentation and modelling of aggregate based on morphological watershed algorithm

为了解决混凝土图像处理中产生新的骨料边缘和边缘移动问题,提出了一种基于灰度图像形态学重构和标记的分水岭分割新算法.该算法结合多尺度理论,采用不同尺寸的结构元素对灰度图像进行形态学重构,然后对重构后的灰度图像进行内部和外部标记,采用形态学极小值标定技术修改梯度图像,最后进行分水岭变换,准确实现了混凝土图像的骨料分割,可用于真实骨料分形几何特性及力学性能研究.运用该方法生成的数值模型进行混凝土单轴受拉数值模拟,验证了该方法的有效性.

In order to solve the problems of aggregate edge producing and edge moving in concrete image processing,a watershed segmentation method combining morphological reconstruction with mark is proposed.Based on the multiscale theory,the method employs different sizes of structure elements to reconstruction grayscale image;then the reconstruction image both internal and external is marked by image pixels and the markers are used to modify gradient image by the way of morphological minimum calibration technique.Finally,the watershed transform of gradient image is performed to achieve the aggregate segmentation of concrete image;and aggregate model established provides a foundation for the research of aggregate fractal geometrical properties and mechanical properties of concrete.The results of numerical simulation under uniaxial tensile test confirm that this method is valid.