Research Of Algorithm For Geometric Surface Reconstruction Based On Point Cloud

The rapid development of both three-dimensional scanning technology and acquisitiontechnology make the three-dimensional geometric model widely applied in game animation,CAD, CAM, reverse engineering and so on. Although, in reality, there is no difficulty toobtain huge and highly precise three-dimensional data information, the construction andprocess of three-dimensional geometric model challenge the processing ability of computer.Therefore, how to rapidly, efficiently and reliably realize the geometric surface reconstructionof point cloud data becomes one of the research emphases and hot topics on computergraphics.Since the surface reconstruction of point cloud data using the traditional triangulationgrowth method inevitably encounters the problem of costing too much time in searching thethird point, which finally leads to low reconstruction efficiency. To attack this problem, thispaper firstly utilizes the adaptive octree partitioning algorithm to divide the whole point clouddata into overlapping subfields, and then makes triangular mesh reconstruction in eachsubfield, which avoids the grid stitching. Secondly, the maximum angle minimizationprinciple has been used to optimize the triangular mesh. Finally, the triangular facetsorientation has been employed to do the consistent treatment on mesh normal vectors. Theproposed method heavily improves the efficiency of surface model reconstruction and theformed grids are able to better show the detail features of models. This paper mainly takesadvantage of both the space partitioning algorithm and the triangular network growthalgorithm to explore the key technology on surface model reconstruction. The main works areas follows:(1) The method of implicit surface reconstruction based on radial basis function is usedto reconstruct the model of point cloud. The algorithm of octree segmentation is used toclassify the point cloud data, then the implicit function based on radial basis function isconstructed to reconstruct the point cloud in each subdomain.(2) Firstly, the adaptive octree partitioning algorithm is exploited to make space division over the whole point cloud data. Secondly, according to the density of point cloud data, thewhole point cloud data is evenly divided to different subfields using the adaptivesegmentation method. Finally, processing on point cloud data coming from each subfield ismade, which greatly raises the computational efficiency.(3) By analyzing the advantages and disadvantages of three algorithms on triangularmesh generation, this paper finally chooses the triangulation growth method to makereconstruction of point cloud model. The three algorithms are method of insertion point bypoint, method of divide and conquer, method of triangulation expanding. By introducing someconstraints, the paper has built triangular meshes with high regular degree.The adaptive octree partitioning algorithm has been used to divide the point cloud datainto overlapping subfields. Then triangular mesh generation is made in each subfield.Experimental results show that the proposed method greatly enhances the efficiency ofsurface model reconstruction. Furthermore, the quality of the constructed grids is also good,which cannot only better reflect the models’ detail features, but also demonstrates the goodrobustness.