University of Twente Student Theses


Sensitivity analysis of rock mass discrete fracture network to UAV 3D point cloud resolution and processing parameters

Mulala, James (2020) Sensitivity analysis of rock mass discrete fracture network to UAV 3D point cloud resolution and processing parameters.

Link to full-text:
(only accessible for UT students and staff)
Abstract:Rockfalls cause a lot of destruction to life and property globally especially in hilly terrains. Photogrammetry plays an important role in monitoring of rockfalls. Both LIDAR and UAVs have been used to monitor the source areas of rockfalls and their trajectories since decades ago. The reduction in the price of drones on the market has broadened the use of the aircrafts beyond the military use. The development of discontinuity set extraction algorithms has also enabled automatic extraction and analyses of the discontinuities in a rock mass using photogrammetry data in slope stability studies. However, the existing approaches use empirical parameter values to collect photogrammetry data and extract the discontinuities from the rock mass whose results are therefore, not transferable. Hence this research aimed to develop a generic methodology of determining the optimal parameter requirements for a UAV flight plan and discontinuity sets extraction from the collected drone data. The study area comprised two slopes located within the Samaria Gorge site in Crete, Greece. Slope 1 is an upper slope east of the entrance to the gorge and slope 2 is road cut of one of the sections of the access road to the Kallergi refugee camp. The automated UAV survey was used collect data from slope 1 while at slope 2, the manual flight was employed. The UAV data was processed using pix4D to generate point clouds. The point clouds were subjected to automatic discontinuity set extraction using DSE. Three main discontinuity sets were successfully extracted by DSE. One of the three structures is horizontal and orienting in the same direction as the slope and the other two are sub vertical in both slopes. The recovered discontinuities were overlapped with the visually identified structures on the RGB point cloud to quantitatively determine the accuracy and effectiveness of the algorithm to extract the discontinuity sets in a point cloud via the computation of classification accuracy indices. The highest overall accuracy results obtained for the slope 1 data were: 40% for knn, 40% for α and 47% for β from the 2cm_GSD_45° point cloud and 33% knn, 37% α and 40% in the β parameter experiments from the 2.5cm_GSD_45° model. Conversely, the highest overall accuracies for the slope 2 data were: 73% for knn, 80% for α and 73% for the β parameter. The accuracy indices results showed that the slope 2 data outperformed the slope 1 data. The poor results in the slope 1 data were mainly due to noise in the data, insufficient exposure of the rock mass surface, too many undulations on the slope surface, occlusions and shadowing effect from the higher vegetation. The optimized parameters based on slope 2 accuracies were: knn 40, α 15 and β 50. The geometric characteristics of the discontinuities in the two slopes were predominantly tiny spacing in the order of 10cm and non-persistent in the order of 0m persistence forming a highly fragmented discontinuity network. In addition to the fragmented geological structures, rockfall triggering factors that are active and frequent in the area accelerate rockfalls down the slopes. Therefore, rockfall mitigations guided by comprehensive hazard mapping are highly recommended. Future, studies are recommended to quantitatively determine the effect of roughness on the extraction of the discontinuities, study the vegetation removal capabilities of the canupo plugin in cloudcompare to maximize its potential in such research work and lastly, future studies can consider checking the influence of other parameters not tested in this research in other terrains with different characteristics and check how the requirements of flight planning and discontinuity set extraction would change or compare with the results obtained in this research.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page