Communications in Information and Systems

Volume 17 (2017)

Number 2

Incremental reconstruction of water-tight surface from unorganized points

Pages: 85 – 113

DOI: http://dx.doi.org/10.4310/CIS.2017.v17.n2.a2

Authors

Rao Fu (Research Center of Biomedical Engineering, Tsinghua University, Shenzhen, China; and Department of Biomedical Engineering, Tsinghua University, Beijing, China)

Cheng Wen (Research Center of Biomedical Engineering, Tsinghua University, Shenzhen, China; and Department of Biomedical Engineering, Tsinghua University, Beijing, China)

Riqing Chen (Research Center of Biomedical Engineering, Tsinghua University, Shenzhen, China; and Department of Biomedical Engineering, Tsinghua University, Beijing, China)

Yifan Fu (Research Center of Biomedical Engineering, Tsinghua University, Shenzhen, China; and Department of Biomedical Engineering, Tsinghua University, Beijing, China)

Jian Wu (Research Center of Biomedical Engineering, Tsinghua University, Shenzhen, China)

Abstract

Most surface reconstruction algorithms require to input all sample points during a single process to construct a surface, but this requirement limits their applications in an incremental surface reconstruction under a consecutive point acquisition process. In this paper, we propose an incremental water-tight surface reconstruction algorithm, which allows the surface to be updated along with the incremental construction of the Delaunay triangulation and guarantee to output a water-tight surface. The core functioning structures, called regular umbrella-covered graph (RUCG) and constrained umbrella (CU), are defined to represent the dynamic reconstruction process. Experiments are carried out to demonstrate the robustness of the proposed algorithm to reconstruct surface from sparse, dense, little noisy point clouds and point clouds of discrete objects. Comparisons with other Delaunay–Voronoi based algorithms are presented to support the effectiveness and superiority of the proposed algorithm. The proposed algorithm is quite simple, fast and non-parametric, and its complexity is only related to the 3D Delaunay triangulation of the points.

Full Text (PDF format)

Research supported in part by Knowledge Innovation Program of Basic Research Projects of Shenzhen under Grant No. JCYJ20160428182053361 and Guangdong Science and Technology Plan under Grant No. 2017B020210003.