Conference material: "Proceedings of the International Conference on Computer Graphics and Vision “Graphicon” (19-21 September 2023, Moscow)"
Authors:Glumova E.S., Serova M.A.
Investigation of Point Cloud Filtering Algorithms for Robust Surface Reconstruction
This paper considers methods of point cloud processing for noise-resistant 3D reconstruction in the context of surface reconstruction under limited computational power. The main causes of point cloud errors and methods of their elimination are presented. The errors in point cloud include noise obtained during object scanning, point cloud redundancy, and visual incompleteness. In the area of point cloud processing, point cloud filtering algorithms are key tools for improving accuracy and reducing noise in the resulting data. An iterative algorithm for filtering out erroneous points using the mean value of neighbouring points' coordinates, as well as a point cloud segmentation algorithm based on the detection of connected regions are considered. It was found that combining several filtering algorithms can lead to better results compared to using separate algorithms.
Three-dimensional reconstruction, point cloud, filtering, polygonal model, SOR, Structure from Motion, Screened Poisson Surface Reconstruction