Digital Rockfall Databases: Developing Best Practices for Semi-Automatic Extraction of Rockfall from LiDAR
In the digital age, the ability to collect and store information of phenomena in our world has become ever more useful, considering the vast improvements of predictive models and simulations. Geohazards are no different; acquiring data in order to characterize, forecast, and simulate geohazards requires the monitoring and documentation of their occurrences, with as much accuracy and detail as possible. Over the previous decade, advancements in remote sensing has resulted in platforms for capturing detailed 3-Dimensional surficial information of our surrounding world, providing opportunities to monitor geohazards with high levels of detail. Systems with oblique views are able to capture near vertical surfaces of rock cliffs, allowing for the monitoring and characterization of rockfall. Much research has thus focused on characterizing and understanding the failure mechanisms of rock slopes, by developing and applying tools for extracting information key to the rockfall phenomenon. Observing rockfall allows for an understanding of its magnitude-frequency distribution, its spatial distribution, its geological and environmental triggering factors, as well as its chaotic interaction with the terrain. The potential power of digital rockfall databases is therefore limited by our ability to extract rockfall from our data – this is the focus of this thesis. Over the past 8 years, semi-automated methods have been developed and improved for extracting rockfall from laser scanning data. Although it is simple in theory, many sequential subprocesses are required, and thus, any errors introduced are capable of propagating, significantly impacting a rockfall database. In this thesis, I aim to build some best practices considering the common tools that have become utilized across different examples of semi-automated rockfall extraction methodologies. I cover my variation of a semi-automatic method for assembling rockfall databases, and refer to its useful application in the domain of quantitative hazard and risk analyses. I discuss a heavily utilized change detection algorithm (multiscale model-to-model cloud comparison, or M3C2) used for isolating changing features in sequential spatial datasets, and discuss its spatial averaging component that has a substantial effect of rockfall extraction. I provide detailed insight into the estimation of rockfall point cloud volumes in 3D using computational geometry-based surface reconstruction methods. I recommend a hybrid surface reconstruction methodology comprised of two methods, the Alpha Solid, and the Power Crust. Together, the hybrid methodology is robust in achieving the correct topology of the rockfalls, while optimally considering prominent concave geometrical features and detailed surficial information. I finish my thesis by discussing the applications of digital rockfall databases into modern quantitative hazard and risk analyses, and I provide recommendations for future work.
URI for this recordhttp://hdl.handle.net/1974/28748
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: