Techniques Of Large-Scale Point Clouds Processing And Surface Reconstruction

With the rapid development of modern science and technology, technology such as numerical control machining, measurement and computer, reverse engineering is playing an important role in industrial design, medical, film, heritage conservation etc. and it is widely used by its special production processes. It digests and absorbs advanced experience and technology to produce new products rapidly based on existing physical models. This thesis focuses on aspects of smoothing, geometric measuring and surface reconstructing on large-scale point clouds in reverse engineering.1. In the aspect of point clouds smoothing, we implement the method of neighborhood filter smoothing for ordinary scattered point clouds based on octree, and propose a method to smooth the millions large-scale point clouds which is attained from projection grating measuring by processing of Gaussian smoothing and bilateral filtering based on the projection grid topological relations of point clouds. The experiment results show that these methods can achieve good effects.2. In the aspect of point clouds measurement, we implement a method to measure the geometrical information of spherical, cylindrical and conical surfaces. The method obtains the geometrical information by solving the quadric equation with meaningful information. The experiment results show that we can gain the accurate geometrical information based on large-scale point clouds efficiently.3. In the aspect of point clouds surface reconstructing, we propose an improved regional growth reconstruction method based on octree. First, find out flat area as the initial growth region, then control the mesh generating process by forced constraint criterion and comprehensive optimality criterion. Additionally, we propose a hierarchical reconstruction method using projection grid. It combines both simplifying and reconstructing operations, and completes the surface reconstruction with Bottom-driven method. The experiment results show that it can reconstruct surfaces from millions point clouds in several seconds and fill in the holes. The efficiency and effectiveness of this method are very satisfactory. Meanwhile, different popular smoothing methods dealing in mesh level are compared.4. We implement a point clouds and mesh processing system to verify the feasibility and practical applicability of those methods proposed in this thesis. It can process and reconstruct point clouds rapidly.Compared to traditional approaches, the different methods for processing and reconstructing of large-scale point clouds which are proposed in this thesis improve efficiency obviously. Next step we will focus on researching a new method by absorbing the advantages of different methods.