Research On Parallelizing Poisson Algorithm Of Implicit Surface Reconstruction

The application of three-dimensional reconstruction in various fields like Virtual Reality and Computer Vision has undergone an explosive surge and abstracted wide attention of research specialist staff.With prevalent use of 3D scanner in recent years,it has been much easier to obtain point cloud displaying features of representation model.However,on t he downside,it accordingly increased the number of point cloud and the complexity of the topology structure.In practical engineering operation and experimental operation,reconstruction of object model for complex topology structure,algorithm will be confronted with various problems,such as poor real-time performance,unfavorable effects,“data blank”,and segment in surface reconstructed.In order to solve these problems,this paper conducted an in-depth research on the reconstruction of point clouds with large-scale complex topology structure,basing on Poisson Reconstruction theory.The main works are as follows:For point cloud data processing,the writer analyzed abnormal points in object model and reduced noises according to the results.Furthermore,point cloud hole was patched by adopting bicubic spline interpolation algorithm,thus eliminating model migration.Then,the least square method is adopted to calculate accurately normal vector of new sampling point.Through the analysis of experimental data,the experimental approach not only filtered redundant data but also preserved the spacial characteristics of object model,which laid foundation for efficient reconstruction of surface later.For the efficiency of reconstruction,regarding large-scale point cloud,poor real-time performance and low efficiency of reconstruction,parallelization basing on OpenMP+CUDA architecture was adopted to accelerate and optimize original processing mechanism of algorithm.And sub-threads allocated by the system was used to complete the operations of each space element.At last,the writer applied parallel acceleration on extracting isosurfaces and octree solving process during Poisson Reconstruction with GPU.Through the analysis of experimental data,it can be found that the experimental scheme has a certain feasibility.Specifically,it not only realized parallel reconstruction of point cloud but also increased the efficiency of point cloud reconstruction,thus enhancing the real-time performance of the algorithm.