Study On The Key Techniques For Semantics-based Description And Modeling Of Complex Parts In Reverse Engineering

As one of the important means of product innovative design, Reverse Engineering (RE) has a broad application background in aviation, aerospace, automobile, shipbuilding, medical and other fields. In these application fields, due to the special external or internal structure, complex parts play key roles on the realization of product function. In general, it is essential for complex parts to experience several prototype design, measurement, modeling and modification process, forming the final products, in which, reverse engineering measurement and modeling technology of RE has made great contributions.In the measurement process of RE, the measurement results of complex parts are often only 3D coordinate information, and a large number of implicit information, such as functional requirements, relevant measurement device information, feature constraint information, functional information, are not directly transferred to the modeling process. The lost information to some extent helps to improve the modeling efficiency and precision. Meanwhile, the main modeling process of complex parts is still in the traditional way of "point-curve-surface". Modeling engineers have to do heavy data processing work to analyze, search, and recovery the geometric topology information, then finish the model reconstruction in the form of "curve-surface-entities". With the increase of part’s complexity, the modeling process becomes extemely time-consuming. In order to improve the efficiency of the reverse engineering design of complex parts, a semantics-based method including measurement and modeling processes is proposed, and the main research contents are as follows:Firstly, the effective implicit information in the measurement process was analyzed and concluded. The semantic information model was established, and the structure semantic data file was designed. A framework consisting of acquisition, parse, process of semantic information and model reconstruction was presented.For multi-view point cloud from same or different sources, registration methods of point clouds under their semantic information were proposed. An edge grouping method was used to obtain the exact position of circle markers, which can be applied to the online registration. Besides, a new method for registration of multi-view point clouds with low overlapping ratios was presented, which can be used to the automatic offline registration of point clouds.In the aspect of noise reduction, to deal with noise in the scattered point cloud and scan line data, two adaptive threshold methods were proposed, respectively, grid-based neighborhood distance and K-means based angular methods. In the process of point cloud simplification, a uniform space sampling method was used to reduce the amount of data and estimate local geometrical characteristics of point clouds. A non-uniform space sampling method with a specified reduction ratio was put forward, which was used in the subsequent interaction reconstruction of curve and surface.According to the characteristics of mechanical parts, an automatic iterative method for data segmentation based on surface fitting error control was proposed. The average intersection angle of normal vectors was used to decide whether point data belong to the same type. Planar and quadric surfaces can be exactly extracted through region growing and error analysis of surface fitting.In order to improve the modeling efficiency under sematic information, both the automatic reconstruction process of non-functional feature and fast interactive reconstruction of functional features were studied. For the sake of satisfying the model reconstruction accuracy of different functional features, a feature parameter optimization method based on a directed constraint graph was proposed. Besides, a B-spline wavelet smoothing method based on error control and boundary continuity was put forward, with the purpose of obtaining higher surface quality.Finally, comprehensive experiments were studied. Both CMMs and laser scanner equipment were used in the surface digitization process. The measurement software MSIS embed with the semantic information acquisition module and RFID wireless module was also developed. The acquisition procedure of semantic information was also designed. MATLAB was the main platform for data reprocessing, and a modeling software prototype named Semantic/RE based on Open CASCADE was developed. Several cases with complex parts were studied to show the feasibility and effectiveness of the proposed methods.