Optimization of the autonomous vehicle-based observation of open pits: rescaling effect of hyperspectral analysis

UAV-based remote sensing has become an essential tool in the mining industry, offering efficient data collection for terrain analysis, mineral exploration, and environmental monitoring. This thesis focuses on evaluating the impact of flight altitude on multispectral imaging resolution and reflectance accuracy, utilizing a DJI Matrice 350 UAV equipped with an Altum PT multispectral camera. Three flight altitudes (60m, 90m, and 120m) were tested over a controlled mining area in Erzberg, Austria, to assess spatial precision and spectral data reliability. The research applied photogrammetry techniques to generate dense point clouds, orthophotos, and digital elevation models (DEMs) using Agisoft Metashape. CloudCompare was used for point cloud segmentation and Python scripting facilitated further statistical evaluations of spatial displacement and reflectance variations. The study examined critical parameters such as ground sampling distance (GSD), spectral reflectance consistency, and geometric accuracy to quantify altitude-related effects on data quality. Results showed that lower altitude flights provided higher spatial resolution but required more processing power, while higher altitude flights covered larger areas at the expense of fine details. Spectral analysis revealed that reflectance values varied across different bands, with near-infrared (NIR) and red-edge bands displaying the highest sensitivity to altitude changes. The comparison between At-Starting-Point (ASP) and At-Ground-Level (AGL) flight strategies demonstrated that AGL flights produced more consistent reflectance results due to a stable sensor-to-ground distance. All analyses were conducted on raw data to ensure precise evaluation before any post-processing adjustments. The research identifies key optimizations for UAV-based multispectral imaging in mining applications, helping to refine data collection workflows.