Research On Mesh Reconstruction Algorithm For Scattered Points

Surface reconstruction is one of the important areas of reverse engineering.The surface reconstruction of scattered data points has always been a main research in function approximation. This paper has studied some related technologies about scattered data surface reconstruction from the actual needs. The main contents are listed as follows:Firstly, a detailed study about space mesh triangulation of the scattered data has been done. Based on the current mesh triangulation method's comparative analysis, aimed at the shortage of other method, presenting an improved triangulation method directly in 3D space. According to the surface shape changing of the corresponding scattered points, a triangular mesh will be generated by using mesh expansion, boundary ring division and close and then contract by the layer. Experimental results show that the improved algorithm can easily deal with a variety of surface space of scattered point clouds and generates the triangular mesh has a fine form, and national layout.Secondly, on the foundation of analysis existing reconstruction method's shortcomings, an improved algorithm that based on the RBF neural network has been presented to construct mesh shape in 3D space. In this section, focused on the following aspects:1. Summed up the RBF interpolation theory and interpolation problem has value's sufficient condition, and have done a detail analysis of the RBF net model based on the theory which can accurate interpolate point cloud data.2. An octree based the rapid k nearest algorithm is presented. Through the establishment of bounding box on point sets, using octree record segmentation process, so that the search of point neighbors is limited to the bounding box of sample points and its neighbor, and through pruning strategies to further narrow the scope of the search, so that the search speed can be well improved.3. An improved algorithm for surface reconstruction based on neural network has been presented. Firstly, making a normalization processing for the cloud data. And then extract the contour lines and segment the surface based contour lines. The method can directly from the Neural Network weight matrix to get the right curve control points or surface control mesh, through weight restriction of Neural Network to achieve the right curve or patch connection. Experimental results show that this method can quickly obtain good shape mesh.