Mobile laser scanning and octree-based statistical inference change detection for geotechnical underground mine monitoring and analysis

Abstract

Convergence and rockmass failure are significant hazards to personnel and physical assets in underground tunnels, caverns, and mines. Today monitoring and analysis of adverse ground behavior are primarily conducted using legacy techniques with low spatial resolution and insufficient data integration capabilities. These limitations create critical risks to underground mine safety and productivity. In three papers, this dissertation explores the potential of Simultaneous Localization and Mapping (SLAM)-based Mobile Laser Scanning (MLS) systems and develops methods for octree-based data processing for rapid geotechnical change detection and integrated analysis of underground mines.

The first paper assesses the quality of SLAM-based MLS data for geotechnical monitoring in two underground mines. Results demonstrated that SLAM-based MLS systems can provide data of sufficient quality to detect geotechnically relevant changes while being significantly more efficient than traditional static systems. Several point cloud alignment strategies were tested and compared. Furthermore, SLAM-specific processing reduced SLAM-drift error and achieved an order-of-magnitude better relative accuracy than quality metrics derived from traditionally deployed tests suggested.

The second paper presents a novel octree-based computational framework for intra-voxel statistical inference change detection and deformation analysis. This method exploits high-density MLS data by maintaining efficient statistical representations of point distributions and performing Chi-Squared tests. The results using three synthetic and field data sets exhibited good performance in detecting rock falls and convergence in real-time. Compared to an M3C2-based method, the developed framework was less sensitive to noisy data, required fewer parameters, and offered significantly better computational efficiency and scalability with large datasets.

Finally, the third paper advances and utilizes the octree-based framework for integrated geotechnical analysis, combining geological, microseismic, and MLS-based change detection data from a third underground mine case study site. The paper demonstrates new capabilities such as geological fault data integration, seismic energy exposure modeling, and combining Random Sample Consensus (RANSAC) classification with Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Statistical analyses revealed correlations among MLS-measured changes, distances to geological faults, and seismic energy exposure. The results suggested that proximity to faults and increased seismic energy exposure could be linked to increased geotechnically relevant deformations.

Overall, this dissertation demonstrates the potential of SLAM-based MLS systems in combination with a novel octree-based data processing framework to enable rapid geotechnical change detection and integrated analysis. These advancements aid in the understanding of complex geomechanical behaviors, ultimately enhancing mining productivity and safety.