Research On Surface Reconstruction And Rendering Based On Scatterd Data

Recently, with the fast and high-precision three-dimensional digital scanning devices emerging continuously, acquiring sampling point sets from object surface becomes relatively simple, cheap and accurate. The process of converting these sampling point sets to geometric model is called surface reconstruction. This technology is widely used in Computer Aided Geometric Design, Computer Graphics, Medical Image Processing, Computer Vision and other fields. According to the sampling point sets acquired by three-dimensional digital scanning devices, this thesis focuses on the limitations of the virtual Operation System for Beautifying Lady's Face and mainly studies on surface reconstruction and Rendering. The concrete research work consists as follows:According to the scattered data surface reconstruction, this thesis refers to the scattered data surface reconstruction algorithms at home and abroad, compares their advantages and disadvantages and draws these algorithms' advantages, and combines with the human face characteristics, and proposes an adaptive implicit surface reconstruction algorithm driven by fitting error. This method adaptively fits scattered data according to fitting errors, improves the algorithm efficiency by reducing fitting times of octree's high nodes, and modifies the weight functions of the partition of unity method and weakens the effect of inconsistency density of the scattered data on surface reconstruction algorithm by adding density weight function. The experimental results indicate that the algorithm's principle is simple and the reconstruction is fast with good effect, which also meets the surface requirement of the virtual three-dimension Operation System for Beautifying Lady's Face.However, when we use the classic marching cubes algorithm for implicit surface triangulation, we find that is the triangular mesh model expands along with their boundary for scattered data reconstructed from open domain. In order to solve this problem, this thesis presents an implicit surface triangulation method with boundary preserved. Firstly, we compute the cubic bounding box of the scattered data and do space subdivision for these bounding boxes, and establish octree storage structure for these scatted data and obtain cubic voxels. Secondly, we choose a voxel which contains scattered data and intersect with aero isosurface as seed and expand along with the zero isosurface. During the expanding process, we use different triangular method strategies for these vovels according to their types determined by the relationship of vovel and the geometric features in scattered data set. The experimental results indicate that the algorithm can produce a triangular mesh which keeps the boundary of the scattered data.