Research On 3-D Reconstruction From Point Cloud

As 3d-reconstruction from point cloud is widely used in engineering, science and art, scholars in the fields of computational geometry, reverse engineering, surface reconstruction and computer graphics focus on searching the 3d-reconstrucion algorithms with low computing complexity and high quality mesh. Our paper studies the topics of mesh reconstruction from point cloud.We do a thorough study on 3-d mesh reconstruction from point cloud with the aim of exploring a less cost algorithm to reconstruct high quality mesh.There has been much work on reconstructing surfaces from point cloud. In the paper we study these work and do a summary for them. After that we choose the Advancing Front Technique (AFT) as our algorithm. We analyzed the algorithm of AFT and improve the algorithm for each step of AFT and the data structure of point cloud.Firstly, with the property of large in scale and non-direct relation between different points, the process of point data and the topology depends on the neighborhood. A great number of operations for query and update are involved to establish the topological relationship. Storage the data points in the octree is a good idea, as this paper does. To establish the octree, the data space must be divided into voxels. The points scatter in voxels. Each voxel maps a leaf node in the octree. The points in each voxel are only storage in corresponding the leaf node. A new approach is adapted to define the neighborhood of different degree for the point in this paper.Thirdly, the second critical step of AFT is how to include a free point into the built mesh to make the mesh is optimal in general. Several guidelines and criterions are proposed. An evaluation function is designed for the guidelines and criterions. The evaluation value for each point in the neighborhood of the front edge by the evaluation function is computed to determine the best point included into the mesh.Finally, the AFT is modeled as an optimization and this can improve the efficiency of surface reconstruction from point cloud. Particle Swarm Optimization (PSO) is used to determine the best triangle, which can make the reconstruction algorithm fast and accurate.The advantages of the improved algorithm prove to be effective by experiment. Our algorithm improves the three-dimensional reconstruction of point cloud data efficiency and accuracy.