Design A 3D Reconstruction System Based On Point Cloud

With the development of game, simulation, graphics technology, and more and more demand for complex physical modeling in industrial design, simulation, 3D film and so on, The attention on research about 3D model reconstruction has been paid more and more. As point cloud data is easy to get, convenient to store, the 3D reconstruction technology based on point cloud has been developed rapidly.A complete 3D surface reconstruction system usually includes the following aspects: first, pretreats the point cloud; second, generates initial triangle mesh from the point cloud; finally, optimizes the initial triangle mesh and gets the final mesh which can be used by other modeling software. This article studies in depth on the three stages:In the point cloud simplification stage, on the basis of discussed the detail of the existing methods, we use the method combined of non-uniform grid and curvature sampling to simplify the original point cloud. We first compute the approximate curvature of every point, and then subdivide the point cloud and make every grid point number no greater than M, then retain some points in each non-uniform grid cell by curvature sampling and the cell size. This will hold more points at the surface region which curvature is bigger, and will not appear holes at flat surface area. And the mesh generation will benefit a lot from this. In the mesh generation stage, we first generate an adaptive ball set to cover the surface of the point cloud, and generate mesh by compute the intersection of the balls. And we control the accuracy of the mesh by controling the generated ball's size to generate different fine degree mesh model. At the same time we also make a regional expansion method based on vertex, which extends a vertex by generated some triangles suround all its sides. We first select a seed point to start the procedure, and at each time we use one point to extend its neighbor points and form triangles until the front point list is null. This method can greatly reduce the mesh triangle's self intersection and holes. In the mesh optimization stage, by defineing mesh optimization operations, each time we select the mesh element with the smallest form factors to execute the optimization operations. This reduces the complexity of the mesh and increases the quality of the mesh. At last we design and implement a three-dimensional reconstruction system of point cloud on the basis of the method introduced in the previous, which can process a variety of different point file formats, and can be enhanced by plugins. Through the testing of different point models, we prove the feasibility and effectiveness of the algorithm.The 3D reconstruction system based on the above approach can be well used in the human body modeling, object modeling, and other game modeling. It is able to quickly and accurately generate computer model from point cloud. It also can be used in other aspects as simulation, medical treatment, cultural relics recovery and etc.