Study On Object Reconstruction Based On Point-cloud Data Via 3D Scanning

In recent years, with the development of computer technology, reverse engineering has become the focus of research and application that it has been widely used in aircraft automobiles, ships, mold and other fields. The key technology of reverse engineering is the object reconstruction based on point-cloud data via 3D scanning. The first requirement of reverse geometry is access three-dimensional point-cloud data of the object quickly and accurately, and then we may do the high-quality surface reconstruction based on the access three-dimensional data.For access point-cloud data, a very important way is to use the 3D scanner. Currently, the mainstream 3D scanner is the optical scanner and laser scanner. To get the further calculation and analysis of the point-cloud data, object reconstruction based on it is the first step. Then through the calculations and analysis on the corresponding three-dimensional model, we may get the geometry, space and other attributes of the subject. The precision and accuracy of the point-cloud data analysis depends on the quality of object reconstruction, so the research on object reconstruction based on point-cloud data has become really a scientific value. This paper proposes a idea and specific method on building the space triangulation. Firstly, choose arbitrary point from { }P = P₁ , P₂ ... P_n as the center of sphere, then establish the optimal sphere with radius r on the center of the sphere c. According to the two-dimensional spherical projection, we may find the subordinate points within the sphere. Loop and when all the discrete points are surrounded by sphere established before, we may get an optimal spherical surface. According to the coincidence of the sphere, analyze the intersection of the sphere and choose the vertices of the point-cloud triangulation. Then connect these vertices to build the space triangulation model. The last step is to fill the small flaw of the space triangulation model, filter non-manifold triangle.After the object reconstruction, we may get the further calculation and analysis of the point-cloud data. And then we can make the feature extraction based on it.This paper proposes a semantic-based feature extraction method. This test classify the point cloud data into three types which are ground,building roofs and the transport vehicles.