Micro-cracks in coal play an important role in the safe mining and coalbed methane extraction processes. The propagation and nucleation of micro-cracks during coal mining are key factors that cause the damage of the coal formation and the overlaying rocks. They are also the main transport ways for the coalbed methane. Therefore, accurate characterization of the micro-cracks in coal is crucial in uncovering the coal failure and the gas transports mechanisms. Imaging methods, such as X-ray CT scan provides an effective way in obtaining the 3D fracture distribution in a coal rock. However, accurate extraction of the fractures in the greyscale CT images remains to be improved, especially for those faint micro-cracks. The difficulties are mainly caused by the limited CT image resolution and the precise segmentation of the micro-cracks influenced by surrounding noises. In this paper, we first used super-resolution and greyscale enhancement techniques to improve the 3D image resolution and quality obtained through the industrial CT scanning. With these applied preprocessing steps, more details of the CT image could be identified. Then multi-scale Hessian filtering techniques was employed to enhance the identification and segmentation of the fractures. Through Hessian filtering, the faint micro-cracks were accurately recognized. Moreover, a connectivity check postprocessing eliminated the noises in the segmented image. The successful recognition of micro-cracks enabled the further studies on the mechanical and transport properties of coal through image-based calculation methods.
Keywords coal resource, Hessian matrix, super-resolution, micro-crack segmentation