Localization of fractures in drill cores using Deep Learning

Abstract: Investigating and assessing rock structures requires extensive manual work, where geologists measure the fractures in drill cores to get a better understanding of the condition of the underlying rock. This study investigates the possibility to use Deep Learning for localizing fractures in images of drill cores, automating and facilitating the work of geologists. The method described is a combination of object detection and segmentation, used to localize fractures and orientation lines. The study focuses on two main aspects: to investigate whether Deep Learning can be used to locate fractures and orientation lines, and whether the data generated by the method can be used to calculate the two angles required to analyze the drill cores, the so-called Alpha and Beta angles. The results show that the localization of fractures can be done from a relatively small data set, with high precision. Angle determination yields poorer results and further work is required to deal with the discrepancies that arise in the segmentation.