ABSTRACT In order to analyze rock slope stability, efficient rock‐mass characterization and 3D numerical modelling are very important. Unmanned aerial vehicle (UAV) oblique photogrammetry, with its low cost, high accuracy, and wide coverage, is commonly used in geological surveys and provides a foundation for rock‐mass quality assessment. Utilizing UAV oblique photogrammetry data, this study proposed a comprehensive workflow achieve efficient 3D mechanical modeling, integrating data collection, rock‐mass structure identification, rock‐mass parameters calculation and numerical modeling. First, oblique photogrammetry was used to gather high‐precision slope images and create a 3D reality model. A semantic segmentation network was then trained to automatically identify rock‐mass structure types. Combined with manually determined discontinuity conditions, the rock‐mass quality of the slope surface can be evaluated using the geological strength index (GSI). After that, the rock‐mass quality within the slope was then estimated using a geostatistical interpolation method based on spatial variability. Rock‐mass parameters were calculated using the Hoek–Brown criterion and represented in a three‐dimensional block model. Finally, through coordinate mapping, these parameters were transferred to a numerical model, ensuring mechanical properties reflect spatial variability and match real‐world conditions more effectively. Each step was validated for accuracy. A case study demonstrated that the heterogeneous model developed using this method outperformed the traditional homogeneous model, providing more accurate predictions of slope failure behavior.
Liu et al. (Wed,) studied this question.