With the development of aviation industry in our country, the application of reverse engineering in CAD/CAM becomes more widely and important than ever. Based on capturing designers' intentions, this thesis studies segmentation of point clouct-and fitting of regular surfaces (including planes, spheres, cylinders, cones, tori, etc.) in structural parts modeling. The main work is as follows: Features and constraints in the reconstruction of CAD models are classfied in detail, and the constraint equations about structural parts constituted by regular surfaces are given. Data segmentation that extracts the edges and partitions point cloud plays an important role in reverse engineering. This thesis proposes an octree-based 3D-cell subdivision method for data segmentation. Firstly, the final 3D-cells are constructed through a refinement process and iterative subdivision of cells. Then, based on normals or curvatures of cells, this method identifies local surface parts through backdating octree. Lastly, feature recognition is implemented automatically according to the sequence of regular surfaces, swept surfaces and free-form surfaces. The least-squares fitting for regular surfaces, such as planes, spheres, cones, tori to 3D data points is researched. The efficiency of computing is improved greatly and the effect of singularities on fitting is avoided.
|