Geometric feature extraction and model reconstruction from unorganized points for reverse engineering of mechanical objects with arbitrary topology

In many industries, the process of creating geometric models of existing objects is an important step in reverse engineering when the prototype is created or modified on the shop floor and when the CAD model does not exist. In this dissertation, the problem of extracting geometric features and reconstructing geometric models from unorganized points for reverse engineering of high precision mechanical objects with arbitrary topology is addressed.; In the proposed research, the object shape is first recovered from its discrete point representation by an interpolating manifold mesh. The mesh surface captures the object topology and linearly approximates the object geometry. In order to recover second order object geometry, a mesh optimization algorithm is proposed to minimize the difference of estimated curvature properties between the original surface and the mesh surface.; The optimized manifold mesh is further segmented to extract the geometric features carrying important engineering information. In the proposed mesh segmentation algorithm, border detection and region growing are integrated together to achieve robust segmentation from noisy data. Border detection process is first applied to extract border regions, which coarsely segment the mesh surface into disjoint patches. The exact border location is then obtained by thinning the border regions into border polygons in the region growing process.; From the segmented mesh patches, the compact and precise representation of feature surfaces is obtained by surface characterization. Based on instantaneous Kinematics, a non-iterative procedure is proposed to efficiently classify regular surfaces and simultaneously estimate their geometric parameters. For sculpture surfaces, novel parameterization techniques are developed to support B-Spline surface fitting from unorganized points.; From the established mesh topology and feature surface representation, the CAD models are automatically reconstructed and represented in ACIS geometric modeler. The reconstructed CAD models can serve as a basis for various design and manufacturing processes. Comparing with other state-of-art techniques, the developed technology in this research has the unique advantage that the process of geometric feature extraction and model reconstruction is fully automated.