Despite several technological advancements, underground mines are still largely relied on visual inspections or discretely placed direct-contact measurement sensors for routine monitoring. Such approaches are manual and often yield inconclusive, unreliable and unscalable results besides exposing mine personnel to field hazards. Mobile laser scanning (MLS) promises an automated approach that can generate comprehensive information by accurately capturing large-scale 3D data. Currently, the application of MLS has relatively remained limited in mining due to challenges in the post-registration of scans and the unavailability of suitable processing algorithms to provide a fully automated mapping solution. Additionally, constraints such as the absence of a spatial positioning network and the deficiency of distinguishable features in underground mining spaces pose challenges in mobile mapping. This thesis aims to address these challenges in mine inspections by optimising different aspects of MLS: (1) collection of large-scale registered point cloud scans of underground environments, (2) geological mapping of structural discontinuities, and (3) inspection of structural support features. Firstly, a spatial positioning network was designed using novel three-dimensional unique identifiers (3DUID) tags and a 3D registration workflow (3DReG), to accurately obtain georeferenced and coregistered point cloud scans, enabling multi-temporal mapping. Secondly, two fully automated methods were developed for mapping structural discontinuities from point cloud scans – clustering on local point descriptors (CLPD) and amplitude and phase decomposition (APD). These methods were tested on both surface and underground rock mass for discontinuity characterisation and kinematic analysis of the failure types. The developed algorithms significantly outperformed existing approaches, including the conventional method of compass and tape measurements. Finally, different machine learning approaches were used to automate the recognition of structural support features, i.e. roof bolts from point clouds, in a computationally efficient manner. Roof bolts being mapped from a scanned point cloud provided an insight into their installation pattern, which underpinned the applicability of laser scanning to inspect roof supports rapidly. Overall, the outcomes of this study lead to reduced human involvement in field assessments of underground mines using MLS, demonstrating its potential for routine multi-temporal monitoring.