A Survey of Surface Reconstruction from Point Clouds

Abstract

The area of surface reconstruction has seen substantial progress in the past two decades. The traditional problem addressed by surface reconstruction is to recover the digital representation of a physical shape that has been scanned, where the scanned data contain a wide variety of defects. While much of the earlier work has been focused on reconstructing a piece‐wise smooth representation of the original shape, recent work has taken on more specialized priors to address significantly challenging data imperfections, where the reconstruction can take on different representations—not necessarily the explicit geometry. We survey the field of surface reconstruction, and provide a categorization with respect to priors, data imperfections and reconstruction output. By considering a holistic view of surface reconstruction, we show a detailed characterization of the field, highlight similarities between diverse reconstruction techniques and provide directions for future work in surface reconstruction.

Real-time scene reconstruction and triangle mesh generation using multiple RGB-D cameras

“…It has been considered difficult to generate meshes on top of MLS processed point clouds. Regarding MLS, Berger et al [7] note that ''it is nontrivial to explicitly construct a continuous representation, for instance an implicit function or a triangle mesh''. Scheidegger et al [46] and Schreiner et al [47] propose advancing front methods to generate triangles on the basis of MLS point clouds.…”

Smooth shapes with spherical topology: Beyond traditional modeling, efficient deformation, and interaction

“…[33], a method is presented to generate a spherical parameterization of a closed surface in the continuous domain by expressing it in a basis of spherical harmonics. A related problem is surface reconstruction from a point cloud [34,35].…”

3D Surface Reconstruction of Noisy Point Clouds Using Growing Neural Gas: 3D Object/Scene Reconstruction

“…In order to avoid non-manifold and overlapping edges, the edge creation step was modified restricting the creation of edges if the winning neurons s 1 and s 2 have already more than input : A point cloud output: 3D mesh 1 For each input pattern presented to the network, the two nearest neurons to the input pattern are selected as winning neurons s 1 and s 2 ; 2 if s 1 and s 2 are already connected by an edge then 3 Set edge age to 0 in order to "reinforce" it; 4 Check edge removal mechanism based on the Tales Sphere; 5 if s 1 and s 2 have one or two common neighbours then 6 foreach common neighbour n i do 7 Create a face f using s 1 , s 2 and n i ; if There exist two neighbours n 1 and n 2 of s 1 and s 2 respectively that are connected and are not common to s 1 and s 2 then 12 Triangulate rectangular hole (Figure 8d): Create two faces using s 1 , s 2 , n 1 and s 2 , n 1 , n 2 ; 13 end 14 else if There exist two neighbours n 1 and n 2 of s 1 and s 2 respectively that are not connected between them and are not common to s 1 and s 2 and. n 1 and n 2 have a common neighbour n 3 then 15 Triangulate pentagonal hole (Figure 8e): Create three faces using s 1 , s 2 , n 2 ; s 1 , n 2 , n 3 and s 1 , n 1 , n 3 ; Edge between n 1 and n 2 is removed;…”

“…the presence of multiple shapes in the point cloud and noise induced by the sensors. Even more recent methods that deal with point clouds fail to provide accurate solutions for some of the aforementioned constraints [3].…”