A new computational approach to cracks quantification from 2D image analysis: Application to micro-cracks description in rocks

In this paper we propose a crack quantification method based on 2D image analysis. This technique is applied to a gray level Scanning Electron Microscope (SEM) images, segmented and converted in Black and White (B/W) images using the Trainable Segmentation plugin of Fiji. Resulting images are processed using a novel Matlab script composed of three different algorithms: the separation algorithm, the filtering and quantification algorithm and the orientation one. Initially the input image is enhanced via 5 morphological processes. The resulting lattice is “cut” into single cracks using 1 pixel-wide bisector lines originated from every node. Cracks are labeled using the connected-component method, then the script computes geometrical parameters, such as width, length, area, aspect ratio and orientation. A filtering is performed using a user-defined value of aspect ratio, followed by a statistical analysis of remaining cracks. In the last part of this paper we discuss about the efficiency of this script, introducing an example of analysis of two datasets with different dimension and resolution; these analyses are performed using a notebook and a high-end professional desktop solution, in order to simulate different working environments.