The significant risk of ground collapse and inadequate ground support in underground mining environments endangers the safety of personnel. To address this risk, Unmanned Aerial Vehicles (UAVs) are among the most suitable tools for use in inaccessible and dangerous areas. In inaccessible mine locations, 3D point clouds can be produced by integrating UAVs with 3D reconstruction technologies, such as photogrammetry. Important geotechnical information can be obtained, including the ability to detect discontinuities, assess rock mass conditions, perform accurate volume calculations, and acquire a georeferenced geometry of inaccessible openings, while remaining outside hazardous zones. Until recently, photogrammetry was the most common technique for the 3D reconstruction of real-world objects. However, there are cases where photogrammetric reconstruction is not feasible due to complex surfaces and poor lighting. 3D Gaussian splatting (3DGS) has emerged as a solution to address these limitations. 3DGS, with its explicit scene representation and differentiable rendering algorithm, enables not only high-quality reconstruction but also real-time rendering. Therefore, the objective of this study is to design a system capable of generating a 3D model by employing a consumer-grade drone equipped with a 360-degree camera to collect high-accuracy data in inaccessible underground mine environments at a reasonable cost. A drone with a 360-degree camera and artificial lighting attached was flown into an underground mine to collect image and video data. The resulting dataset was used for 3D reconstruction using both 3DGS and photogrammetry, and the results demonstrated that 3D models can be successfully reconstructed using this system in underground inaccessible areas.
BINALA et al. (Thu,) studied this question.