The Research On Simplification Of Scattered Point Cloud And Triangle Mesh Reconstruction

Reverse engineering was widely introduced into the research and application ofmodern science and technology, such as computer graphics, game development,simulation and many other areas. For the increasing requirements of reverseapplication, completing the reconstruction of the model efficiently and accuratelybecomes a hot topic. The simplification for point cloud and the reconstruction oftriangular mesh are the two main research emphasis of this thesis. And the thesis isorganized as follows:1. Discrete point cloud is the major research object of this thesis. Afterdiscussing the principles and scope of the existing algorithms for constructingtopological relations, we proposed a grid strategy to divide the space into quadraticspace for the searching of the K-neighborhood of target point.2. On the basis of discussing the principles and characteristics of variousstreamlined algorithms of point cloud, we introduced an improved method to solvethe disadvantages of the single streamlined algorithm. This method can be dividedinto three parts, firstly calculating the geometry element such as surface normalvector and curvature, and then dividing the data point into two categories: flat pointand high curvature point, finally using simplification method with feather reserved intypical region and simplification method with uniform effect in flat area toaccomplish the modest streamlining of point-studded. The algorithm can be adaptedto different occasions by changing the threshold. And experimental results showedthat it can effectively preserve many characteristics of the model under certainrequirements of the streamlining.3. After analyzing the triangulation of point cloud in this thesis, we put forward amultiple constraints plan on the regional growth triangulation algorithm. Thisalgorithm is an effective solution to the shortage of high complexity in the pasttwo-dimensional division. Starting with the seed triangle, the best point is selectedfrom those points which are near the both ends of active layer which has beenestablished in edge list. Combined with dihedral angle and the used point first strategies, more and more triangle has been created. In addition, it is likely toaccelerate the detection about expansionary of the side by establishing adjacent table.Different types of surface construction can be composed of these basic processes.Experimental results showed that the algorithm can generate clear grid model.In this thesis, these algorithms have been designed in VC++6.0with OpenGL.In support of this software we can use the basic operating functions of point cloudmodels and experiments to implement streamline algorithm and the gridding method.