Research On Key Technologies Of Automatic And Accurate Surface Reconstruction In Reverse Engineering

Reverse engineering, which is an important technology of product creative development and rapid manufacture, has become very popular in recent years. It has been widely used in aerospace, automobile, mould and medical fields. As the key technology of reverse engineering, surface reconstruction is always the hot topic for researchers. Due to the development in measurement techniques and computer science, rapid progress has been made in surface reconstruction. But for its theory and applications, there are still some issues need to be resolved. For example, the automation level of surface reconstruction is low and the accuracy loss of the reconstructed model may be great. Or the model reconstruction and its applications have not been fully integrated. To improve the above situations, several key techniques of surface reconstruction are systematically studied and implemented in this thesis. The main topics are as follows.The geometric attributes, i.e. normals and curvatures, are the basis of surface reconstruction. In order to improve the accuracy of curvature estimation for the unorganized point cloud data in reverse engineering, a moving-least square (MLS) surface based method is proposed. Analytical close form solutions for principal or normal curvatures computation are derived from the implicit definition of the MLS surface. As the exact curvature is dependent on the Gaussian factor used in the MLS surface definition, an empirical formula for determining the appropriate Gaussian factor is given.Segmentation is an important intermediate process in reverse engineering. In order to get meaningful and well aligned segment patches with minimal user assistances, this paper develops a Morse-Smale (MS) complex based segmentation method. The presented algorithm automatically segments the triangular mesh by constructing and simplifying the MS complex. Furthermore the automatic separation of the transitional features, i.e. fillets and vertex blends is realized, which is helpful in high-quality surface approximations and extraction of advanced features, e.g. machining features.The feature-based surface reconstruction strategy has the potential to catch original design intents and thus improve the quality of the reconstructed models. The critical techniques in the feature-based method are feature recognition and parameters extraction. A method to automatically identify the surface types is presented. A statistical algorithm is used to reduce the effect of outlier points and computation errors. For the parameter extraction of extrusion and rotational surfaces, a refinement algorithm using point registration is presented to improve the accuracy of the computed parameters.The complex free-form region needs to be subdivided first and then each sub-region is fitted by B-spline surfaces separately. In order to help determine the numbers and locations of the B-spline patches, a region growing method is presented to subdivide the free-form surfaces. The seeds are automatically created by clustering the points with similar curvature values. For each seed region, the algorithm iterates between region growing and surface fitting to maximize the number of connected vertices approximated by a single Bezier surface.Direct slicing of point cloud is an effective approach to integrate reverse engineering and rapid prototyping. Since direct slicing methods take discrete point data as input and connections between the points are absent, great accuracy loss or structure faults may happen to the layered model. This paper presents a method to extract the topological structure from the point cloud and applies the structure into the direct slicing process. With the extracted topological information, the contour structures can be determined as prior, so the slicing process can efficiently and correctly handle complex shapes with multiple contoured slices. It also helps preserve important features and reduces errors caused by topological changes. A curvature adaptive method is used in 2D contour and layer thickness determination, which can balance the machining accuracy and efficiency in rapid prototyping.With the study of above key technologies and relative algorithms, a surface reconstruction system is developed by using C++ and OpenGL. The feasibility and validity of the systerm and relative algorithms is verified by some typical parts.