Study On 3D Surface Reconstruction For Point Cloud Data

In the modern industrial design and manufacturing, it is often required to digitize existing parts and establish its mathematical modeling. Collected by measuring equipment, the information obtained with the surface of spatial data points, or point cloud models. Point cloud model has many advantages, for example, a simple data structure, compact storage space, strong ability to express details. For different needs, often using different methods of surface expression of the models they represent surface reconstruction, this approach is called reverse engineering. With the performance improvement of measuring equipment, Point cloud model has large-scale, high-density characteristics, it has higher requirements in the point cloud model surface reconstruction algorthm's efficiency and performance.Point cloud surface reconstruction can be devided into two categories: explicit and implicit. Explicit surfaces prescribe the precise location of a surface while implicit surfaces represent a surface as a particular isocontour of a scalar function. Popular explicit representations include parametric surfaces and triangulated surfaces, field function method and RBF method are widely used in implicit surface.The main tasks of this paper are following:1. We study of three kinds of typical surface reconstruction algorthm ofexplicit and implicit methods, they are Power Crust algorithm, distance function method and radial basis function variational reconstruction method, we also analysis deeply algorithm theory, complexity and scope. In this paper we mainly discuss RBF implicit surface method.2. The discussion of the principle of RBF interpolation, stability, uniqueness and solution methods provides a theoretical basis for this algorithm. We sum up four ways to improve the efficiency of RBF to solve and deal with large-scale point cloud data, namely the use of compactly surrprted radial basis fuction, multi-level method, partition of unit and multipole method.3. We proposed an adaptive center selection reconstruction algorthm. We use k-d tree to establish point cloud data structure. In the local area of center point, using quadric surface to approximate point cloud data, determing the influence radius through point density and surface geometry. At last, compactly supported radial basis function are used to get the global solution, surface following method is used to exact triangle mesh and output results.Our Algorithm does not need to add additional constraint points and solve quickly. In this paper, we use C++ language to implemented our algorithm. Experimental results show that our algorithm is adaptive, fast, robust, etc, good results can be obtained for both ideal point and density irregular point data.