Xu, ChaoshuiDowd, PeterYazdanpanah, Mohammad2025-07-282025-07-282025https://hdl.handle.net/2440/146361The use of remote sensing tools - specifically laser scanners and photogrammetry – has gained popularity in mining and geotechnical engineering. They allow us to explore various aspects of a rock in detail, from its geological structures, weathering, discontinuities and surface coarseness to its spatial arrangement of different features. In this regard, the development of 3D digital technology, such as laser scanning and more advanced digital cameras, has helped to acquire more accurate information about the 3D structure of rocks more safely and efficiently. The development of associated data analytics has further allowed us to gain more thorough and accurate geometrical and geological insights within a short period of time. Assessing tunnel and slope stability depends critically on the prior knowledge of geological discontinuities. These discontinuities act as weak planes within a rock mass where unstable behaviours occur, which might lead to collapses of the rock excavations. Therefore, in engineering applications, it is critically important to determine their geometric attributes, including their number, position, orientation, persistence, spacing and roughness. These attributes are usually obtained from the judgment of engineering geologists and are based on rock mass engineering principles, thus requiring more accurate, repeatable and objective approaches. To address some of the issues mentioned above a reliable and robust procedure was developed to identify the discontinuity surfaces of rocks with a high degree of accuracy from 3D point cloud data. A new method is proposed to segment surfaces based on image processing by taking into account both geometry and colour information. The geotechnical behaviours of rock discontinuity surfaces are then investigated to understand the impacts of different surface roughness on the shear strength. These procedures lead to the proposal of a new model for the shear strength of rock discontinuities, which incorporates the impacts of different surface roughness components. This research contains three main components: The first part of the research deals with the automated detection of rock surfaces and their features from unstructured 3D point cloud data obtained from a laser scanner or photogrammetry. This is a significant step in facilitating more advanced approaches in this area. The manual detection method has gradually given way to more automated methods due to issues such as cost, efficiency, risk and subjectivity. Given the advances of surveying technologies and the complex nature of rocks caused by weathering and surface alterations, there is a need for an automatic and robust method to extract features from unstructured three-dimensional measurement points. The method developed in this work includes the following steps. Firstly, normal vectors for each point in the point cloud data are calculated, together with the curvature, which measures smoothness. Then, an innovative "region growing" approach is developed. The method is based on a step-by-step growth of flat surfaces within a point cloud, where new points added to the surfaces must satisfy a neighbourhood alignment condition. A threshold is imposed on the process to ensure that only points lying on nearly the same plane and with a normal vector nearly parallel to one another are added to the surface being constructed. The method has been applied to two unstructured point cloud datasets of natural rock surfaces, which has demonstrated its capabilities for practical, reliable extraction of rock surfaces from massive, disorganized point clouds. The second part of this research has further developed the method outlined above by incorporating the colour information of the point clouds. Laser scanners and photogrammetry can now capture RGB colour information together with the coordinates of 3D points. An algorithm is proposed in this research to segment point clouds based on colourimetric and spatial proximity. The algorithm is based on two image-processing techniques: phase analysis and tensor voting. Phase analysis uses the phase congruency approach, which is a frequency-based method to process images for the detection of surface edges or boundaries. To derive a more comprehensive 3D structure of the rock surface, tensor voting is then used to explore points with their immediate neighbourhood to inform each other to conform to the continuity and proximity constraints. This interaction helps each point to measure its structural significance. By using these two techniques, discontinuities can be more accurately identified, and the segmented patterns are then projected onto 3D point clouds to produce a more accurate and detailed mapping of 3D rock surfaces. In the third part of this research, the impacts of rock surface roughness components on the geotechnical behaviours, particularly the shearing resistance, are investigated. It is well known that rock joint shear resistance is strongly influenced by the roughness of the joint surface. However, the quantification of these influences is a very challenging task, and this remains a very active research topic in this area. In the approach used here, joint surface roughness is analysed using the Fourier Transform decomposition to define different roughness components. Two different shearing mechanisms are then used to describe the shearing behaviours of the joint, namely asperity sliding and asperity shearing failure. A threshold curve is proposed to distinguish these two shearing mechanisms based on surface geometry, rock properties, and applied normal stress. Despite much research in the past, existing models fail to apply these mechanisms adequately for rough joints. Based on the approach used here, a new shear strength model for rock joints is proposed and the ability of the model to estimate joint strength is validated by numerical shearing tests conducted using the Finite Element Method under different conditions.enRock massPlanes of weaknessRemote sensingLaser scannerPhotogrammetryImage processingPoint cloud processingSegmentationFourier TransformShearing mechanismsThesis