Distributed Surface Reconstruction

Patrick Komon

Supervisor(s): Projektass.in Diana Marin, BSc MEng

TU Wien

Abstract: As 3D scanning technology is advancing, both quality and quantity of available point cloud data is increasing. Many applications require the reconstruction of the surface of a scanned object as a 3D model. As scans become exceedingly detailed, point clouds become larger and surface reconstruction more computationally challenging. A fast and scalable solution for the reconstruction problem is needed. We constructed and implemented a scalable distributed surface reconstruction algorithm called DistributedBallFilter based on the recently developed BallFilter algorithm. We executed our implementation on the VSC3+ high performance computing cluster and empirically analysed its speedup as well as its parallel scaling behaviour. In our tests for DistributedBallFilter we achieved running times around five times faster than with BallFilter.
Keywords: 3D Reconstruction, Computational Geometry
Full text:
Year: 2023