A RANSAC Based CAD Mesh Reconstruction Method Using Point Clustering for Mesh Connectivity

Mesh reconstruction is widely used in industrial reverse engineering and other 3D scanning applications. Many of the methods used can accurately and swiftly reconstruct a model from a given input scan dataset – however, in some applications, simplification of the output mesh may be needed to render further computation tractable. This paper presents a method based on a recursive RANSAC point removal approach, where an input point cloud is deconstructed into a set of convex point clusters, each corresponding to a mesh face of the final output model. The combined use of cluster analysis and RANSAC model extraction allows the method to operate on a relatively small number of possible vertex candidates for the final mesh, leading to an output mesh with relatively few faces compared to other approaches.