Researchers at Seoul National University of Science and Technology (SEOULTECH) have developed a machine learning-based method to improve geological assessments of rock faces.

The technique, called Roughness-CANUPO-Dip-Facet (R-C-D-F), enhances the accuracy of measuring dip angles and directions by identifying joint embedment points, key features in rock structures. This fully autonomous approach could improve precision and safety in large-scale construction projects, including tunnels and mines.

Machine learning has been increasingly applied across scientific disciplines, including geological engineering. Determining the dip angle and direction of rock facets is essential for ensuring structural stability in underground construction. However, current machine learning models often struggle to differentiate between joint bands – broader, less distinct areas within rock – and joint embedment points, which are more precise indicators of surface orientation.

To address this issue, a SEOULTECH team led by Professor Hyungjoon Seo developed the R-C-D-F method. The multistep process uses filtration techniques to remove joint bands while retaining joint embedment points, improving measurement accuracy. The research was published in Tunnelling and Underground Space Technology.

The method begins with a roughness analysis of a 3D point cloud from the rock surface, removing minor irregularities and noise. Next, the CANUPO algorithm classifies geometric characteristics to isolate key features. A further filtration step eliminates connecting rock segments based on dip angles, ensuring precise measurement of each rock section’s orientation.

Tests on real tunnel face images showed accuracy rates between 97% and 99.4%, with 100% of joint bands successfully removed while preserving 81% of joint embedment points. The method operates without human intervention. “By automating the process of filtering and segmenting rock features, it reduces human error and computational inefficiencies, making it ideal for modern infrastructure projects that demand high accuracy and reliability,” said Prof. Seo.

The researchers believe the R-C-D-F method could have broad applications in structural and geological engineering.

“The R-C-D-F method’s integration of ML and deep learning ensures reliable and accurate geological data processing, which can directly improve the safety of large-scale engineering projects like tunnels and underground structures,” Prof. Seo added. “It could also enable the development of smarter and faster geological analysis tools, reducing costs and improving efficiency in industries reliant on subsurface exploration and infrastructure development.”