Multi-scale voxel-based algorithm for UAV-derived point-clouds of complex surfaces

Point-clouds acquired using remote sensing, such as lasers or photogrammetric techniques, have grown exponentially in the last decade, providing data containing hundreds of millions of points with a density reaching hundreds of thousand points per square centimetre - when working on relatively small areas. This progress has been developed concomitantly with an increasing democratization of UAVs, in such a way that even in hardly accessible areas UAV-based photogrammetric techniques have provided researchers with detailed point-clouds. Provided those dataset, the paradigms have therefore shifted from data collection difficulties, to data processing limitations and usage. The present contribution therefore aims to provide an example of data processing using a voxel-based approach of gravel surfaces in order to derive more than the traditional surface roughness. The algorithm decomposes the surface into voxels and sub-voxels in order to determine if there are surfaces overlapping each other, showing the presence of cavities. Such dataset will then provide the necessary dataset to create metrics to improve the modelling of boundary layers dynamics in computational fluid dynamics for instance.