Accurate identification and segmentation of rock discontinuities are essential for geotechnical and geological research, especially for slope stability analysis and construction planning. Traditional field methods are time-consuming and prone to subjectivity. Remote sensing technologies such as LiDAR and photogrammetry produce high-resolution datasets that, when combined with deep learning methods, enable efficient and objective rock surface analysis. This study presents a comprehensive approach to detecting and classifying rock discontinuities using deep learning models applied to 3D data derived from LiDAR and photogrammetry. We evaluate the performance of Convolutional Neural Networks (CNNs) and image segmentation architectures like U-Net in classifying surface features and distinguishing planar and non-planar discontinuities.
Ghazaryan et al. (Thu,) studied this question.