The Research Of Triangulate Surfaces Division Algorithms Based On Depth Image During The Process Of 3D Reconstruction Gao Yining , Han Xie

With the development of computer graphics technology, especially the atte-ntion for three-dimensional reconstruction, making the triangulated surfaces reconstruction of the scattered point cloud data collection and the Optimization problem after reconstruction, has become a hot research topic.And Research results have a very important practical value in virtual reality, model reconstruction etc.In this paper,the three-dimensional reconstruction is based on the principles of binocular vision. Humans using their eyes to get the outside information, processing, storage and understand them.Computer vision theory is based on the principles of human vision,By using the binocular camera to mimic the human eye to get the perceived image information.Then the computer imitate the human's brain ,conversing the image to digital signal, calculation and processing it.As to get the target object's three-dimensional information,and reconstruct it,like people to see through the eyes to obtain the three-dimensional world.The main works in this dissertaition are as follows:1,Describing the experimental principle,the method of obtain the target image,the calculating of the datas and their storage method. Using the binocular vision theory to obtain three-dimensional point cloud, and store the three-dimensional point cloud in the octree ;2,Introducting several classified method of triangular patches on the scattered point cloud.Including:the triangulation based on Delaunay triangles and Voronoi diagram,Crust algorithm,zero-set method,α-shape algorithm,and the RBF(radial basis function) algorithm used on the reconstruction of the scattered point based on implicit surface.And describes the algorithms'basic idea in details;3,The core of this paper is presented the method of based on the growth of neurons GCS neural network,and applied this method to the reconstruction of the scattered points cloud.Learning algorithm of the GCS and its characteristics.Using it to train the scattered point cloud,making the last patch reconstruction have a better approximation of the point cloud,maintain the topology datas,and also make the reconstruction more smooth;4,Three optimized method of triangular patches surfaces are introducted after the reconstruction.And the methods used in this paper is splitting the points to lower the degree.To ensure that the results are more apporximate to the target object;5,This experiment is use of a binocular camera to get a box's picture in 45degree.According to the methods we described in this article to get the point cloud data,and use the triangulated surfaces to reconstuct the point cloud.The comparison algorithm is Crust algorithm which is used in three-dimensional reconstruction.