Research And Realization Of Three-dimensional Reconstruction Based On Discrete Point Cloud Data

With the development of depth camera, laser ranging, CT, MIR and others of three-dimensional measurements and information acquiring technology, it is possible to include much information about discrete data of measured object surface. Therefore, three-dimensional reconstruction based on discrete data has become the focus of research in the field of artificial intelligence and it has wide range of applications.Discrete data often appears in the form of point cloud. The traditional three-dimensional reconstruction methods are based on the object surface of certain characteristics which have been designed, those usually have some limitations. So, this thesis discusses three-dimensional reconstruction common methods based on simple point cloud and complex point cloud.First, this thesis preprocesses data of three dimensional point cloud and displays directly simple point cloud in the VC + DirectX platform, thus achieves three-dimensional reconstruction based on simple point cloud. Then this thesis researches triangulation and three-dimensional cut in the space based on complex point cloud. An improved triangulation algorithm to complement the drawbacks of traditional algorithm is proposed when there is a non-convex set of points, or inner rings, the probability of triangulation is easy to cause errors, as well as to avoid a phenomenon of the two triangular pieces with "total circle", caused by just deleting the one of them, and then reflects the topological connections among discrete points of object surface. The proposed approach reveals effectively surface of the shape with discrete data.In addition, to rebuild a vivid object, this thesis also discusses the rendering technology based on the grid, namely, to add lighting and materials processing, and texture mapping to triangular mesh, which can improve the rendering speed and rendering effects greatly.Finally, after the combination of the above steps, this thesis establishes a system platform. A new data format based on complex point cloud and a data frame structure in terms of developed platform are proposed, which is used to store the data of three-dimensional reconstruction. Correspondingly, three-dimensional grid data file is established and loaded in VC + DirectX platform. Finally, three-dimensional reconstruction experiments of two different kinds of point cloud are carried out. The experimental results indicate that the proposed method is suitable for most objects and is universal, simple, low cost, high precision, strongly practical usefulness, compared with traditional three-dimensional reconstruction methods.