Study On Segmentation And Constraint-based Feature Reconstruction In Reverse Engineering

With the developing of computer and measurement technique, reverse engineering has become a technology for product innovation instead of traditional methodology for product copy. It has been widely used in rapid development of new products such as automobile, motorcycle, airplane and mould die etc. During the process of product development based on an existing physical part in the absence of the original design information, the reconstruction of CAD model is the most important step. In order to generate CAD model that more accurately represent the original design intent, a new systematic scheme for parametric feature CAD model reconstruction from unorganized 3D points is proposed. The related key techniques including geometric feature, constraint and parameterization are researched intensively.The basic theories including the definition, classification and representation of geometric feature and constraint are presented firstly. The features in reverse engineering are distributed into three feature spaces including bottom extraction feature space, up combined feature space and B-rep/CSG unit feature space. The feature element composition of each feature space and the mapping relationship between feature spaces are also investigated.In the process of generating a CAD model from point cloud data, the segmentation that extracts the edge features and partitions the 3D point data is indispensable. In this research, a new efficient segmentation method was proposed without using triangulation of unorganized points. Based on the subdivision of enclosure box of point cloud and the change of scattered points' curvature, edge feature points are extracted and point cloud data is partitioned into multi-patched surface regions.With respect to the reconstruction of sweeping, revolving, extruding and lofting features, a practical re-parameterization method based on constrained reconstruction of sectional combined curve feature is proposed. The reconstruction process is made up of three steps. Firstly, section data is acquired using point cloud slicing method. Secondly, plane discrete curve is subdivided into feature segments based on region growing method. At last, the sectional combined curve feature is reconstructed by simultaneous multiple curves fitting to segmented curves data. While the sectionalcombined curve is acquired, the parametric feature model can be rebuilt using conventional forward design method.Quadric surface features are abundant in versatile industrial products, and geometric constraints such as perpendicularity and parallelism etc. are often satisfied between them. How to recognize the quadric surface feature and reproduce the geometric constraints are important for accurate reconstruction of CAD model from point cloud. Here presented a new region growing method to segment quadric surface data from 3D scattered points. The major novelty of this algorithm lies in the initial value selection for region growing and the new criterion application for judging point connectivity. After that, a general mathematical model for simultaneous multiple surfaces fitting subject to geometric constraints is established. The regularities and symmetries required by engineering applications are reproduced in CAD model reconstructed based on parameters obtained by resolving the optimization model.The key technologies and algorithms presented above are all implemented in the reverse engineering CAD modeling software RE-SOFT. Experimental results show that the application of them can evidently improve the precision and efficiency of model reconstruction.In the end, a number of advanced topics, concerned in reverse engineering field in future, are addressed.