Automatic Extraction of Surface Structures in Digital Shape Reconstruction

One of the most challenging goals in digital shape reconstruction is to create a high-quality surface model from measured data with a minimal amount of user assistance. We present techniques to automate this process and create a digital model that meets the requirements in mechanical engineering CAD/CAM/CAE. Such a CAD model is composed of a hierarchy of different types of surfaces, including primary surfaces, connecting features and vertex blends at their junctions, and obey a well-defined topological structure that we would like to reconstruct as faithfully as possible. First, combinatorially robust segmentation techniques, borrowed from Morse theory, are presented. This is followed by an algorithm to create a so-called feature skeleton, which is a curve network on the mesh that represents the region structure of the object. The final surface structure comprises the optimally located boundaries of edge blends and setback vertex blends, which are well aligned with the actual geometry of the object. This makes the surface structure sufficient for an accurate, CAD-like surface approximation including both quadrangular and trimmed surface representations. A few representative industrial objects reconstructed by Geomagic systems illustrate the efficiency and quality of the approach.