Surface reconstruction from unoriented point clouds by a new triangle selection strategy

Abstract

This paper presents a surface reconstruction algorithm that takes an unoriented point cloud as input and produces an interpolating surface in the form of triangulation. Based on region-growing and Delaunay approaches, this algorithm aims to address the difficulties of reconstruction from point data with imperfections. Starting with a seed triangle from the Delaunay tetrahedron result for input points, the surface is gradually formed by adding the linked Delaunay triangle from the current boundaries one by one. During surface growth, the topology errors and the quantity of the holes generated by adding inappropriate triangles can be reduced by changing the triangle selection criteria and adjusting the addition order of the triangles. We evaluated our method using a wide range of datasets, and this method compares well to popular classic and current algorithms with unoriented input points and triangulated surface output. In addition, to achieve results with a small number of holes on the generated surface, a detection and repair approach is proposed to turn the holes of various shapes into smooth surfaces.