Research on the semi-automatic fracture identification for rock quality designation determination

Abstract

The study of semi-automatic fracture identification for Rock Quality Designation determination (RQD) has been carried out because, in recent years, deep learning has gone deep into engineering geology. Computer technology can help people finish basic tasks like Rock Quality Designation calculation. The objective of this dissertation was to use the Holistically-Nested Edge Detection[48] (HED) method to automatically identify fractures in rocks and return Rock Quality Designation (RQD) values after adding human judgment. This dissertation mainly carried out the following research work: 1. Image preprocessing, adjusting image brightness, and removing image noise can improve image quality. 2. Holistically-nested edge detection, use the trained model to obtain the images after Holistically-nested edge detection. 3. Matlab image processing, combine the processed image with human judgment, and output the automatically calculated Rock Quality Designation (RQD). Combined with the HED method of deep learning and human judgment, the machine can find the detailed features of the rock core images in a large amount of data, and then judge the edge according to the features, which can well deal with complex rock core images and is of great significance to the edge detection of this field.