Research On The Surface REconstruction Based On Geometric Shapes

While free-form surface reconstruction is an important technology in reverse engineering which can realize digital duplication and improvements, it has a wide prospect. Usually, surface reconstruction adopts the basic mode of point,curve,surface but it is complicated. Another method is on the basis of triangle meshes segmentation with the disadvantage of little precision. Then direct surface reconstruction using geometric shapes by triangle meshes is arised, and it can avoid the point segmentation and curves fitting.In order to get a large number of point data and improve the measurement speed, normally non-contact measurement is used, such as a laser scanner. First, this paper studys the theories, methods and the characteristic of existing measurement machines.By numbers of models measurement, the research is based on the point cloud acquired by TDV800 laser scanner in this paper. Pre-processing which includes format conversion, data merge, noise filter, coordinate transformation and data reduction is observed and proved by physical models, mainly to offer points data with reliable quality.In the process of geomatric shape construction, two basal but important problems are researched such as convex hull and Delaunay triangulation which are used to subdivide a point set. This paper presents region segmentation by four maxima and minima points, which are saved in the link list of convex hull vertex. Meanwhile, these points divide other points into two areas, only to consider the outer points since the inner points mustn't be convex hull vertexes. The new areas are subdivided untill all the areas are NULL by the farthest distance method, in which we can get all the convex hull vertexes and save them in the link list. These points are sorted and connected by anticlockwise vector angles. Meanwhile, it can judge those non convex hull vertexes. According to the property of maximum the minimum angle and four points of a circle presented in this paper, the topology annexation is constructed with point by point scanning, which accomplishes the Delaunay triangulation construction.The paper adopts surface reconstruction by a shape. First, triangle mesh is stored and called by STL. The simplexes are classified and computed in order to obtain a interval by the cicumsphere radius and centre of every simplex. Data structure is designed to save them. Through self-selection, a-shape can be reconstructed by different a parameters.α-shape avoids cockamamie calculations and accelerates the speed. The paper usesγshape in consideration of non-uniform samples by modulating the global factor a with a local scale factor, which is called y. This paper determines the value of gamma by using the distance from a point to its negative pole. To solve holes and bridges, the minimaγof its every vertex is identified as the local scale factor for a simplex, which is used to reconstruct a surface.Fouse on the research contents, a visual software of surface reconstruction based on gemetric shapes is designed by using VC++ programming and OpenGL graphics. The interface design and operation of this software are demonstrated by existing point cloud samples, which realizes the theory in this paper. The data provided by the Stanford 3D Scanning Repository is reconstructed, such as Dragon, Happy Buddha, Stanford Bunny.