Assessing the Feasibility of Photogrammetry for Underground Mine Monitoring: A Simulation-Based Study

Abstract

Underground mining operations face significant challenges in monitoring structural stability and safety due to limited visibility and access. This study evaluates the feasibility of photogrammetry for underground mine monitoring using a simulation-based framework, bypassing resource-intensive field experiments. Synthetic datasets, generated via Blender and MATLAB, were used to assess photogrammetric accuracy, scalability, and limitations under varying lighting (25–100 lux), noise, and dust conditions. Three Structure-from-Motion (SfM) algorithms—COLMAP, OpenMVG, and Agisoft Metashape—were compared to evaluate performance robustness. The results indicate a Root Mean Square Error (RMSE) of 2.3 cm for COLMAP at 50 lux, improving to 1.9 cm at 100 lux, with OpenMVG and Agisoft Metashape showing comparable accuracy (RMSE 2.1–2.5 cm) but varying processing times. Challenges persist in low-light conditions (RMSE 3.8 cm at 25 lux) and complex geometries, exacerbated by simulated dust (RMSE 2.5 cm), highlighting the need for supplemental lighting and advanced image preprocessing. Point density ranged from 8,500 to 12,300 points/m², with dust and noise reducing detail retention. These findings suggest photogrammetry’s potential as a cost-effective, non-invasive monitoring tool, complementing traditional methods like Light Detection and Ranging (LiDAR). The simulation-based approach provides a scalable framework for optimizing photogrammetric workflows, supporting Industry 4.0 trends in mining safety and efficiency. Future work should validate these results in real-world conditions to address environmental variables like dust and humidity.