The Research On Post-Processing Technology Of Large Scaled Scattered Point Cloud

With the development of laser-based scanning technology,the 3D coordinates of an object can be accessed rapidly with high accuracy.However,quantity of the redundant 3D surface data are so more that it is named as "point cloud".To process these point cloud effectively and to generate a CAD file that can be used in manufacturing,machining or reproducing system,is one of the core mission of reverse engineering.Based on that background,a series of key technologies in point cloud post-processing are studied in this paper.Firstly,the redundant point cloud with scattered points should be sampled before further processing.To sample point cloud,the topological relation of points must be reconstruction.This is done by finding the k-nearest neighbors of a given point.Once the topological relation is constructed,one can calculate the normal,curvature, surface vary et al parameters,which define the local feature of the surface hidden in the point cloud.Then,based on local feature,point cloud simplicity is easy to be done.Secondly,quadratic surface often appears in traditional work-pieces.But unfortunately,there is no linear fitting method for cylinder and cone.By means of the concept of 'Faithful Distance' and Non-linear Least Square approach,a novel non-linear fitting of quadratic surface algorithm is presented so that the initial value can be chosen rationally.Thirdly,a new way of Triangulation is proposed.Thanks to the high density of point cloud,our method can be implemented with a step out strategy.All the vertexes of the result triangular mesh model are from the original point cloud.Also,the mesh can represent object with different resolution.Besides,one can control the distance between mesh model and original points.Finally,to reconstruct 3D freeform surface,B-Spline algorithm is developed.A topological grid data can be formed by employing of ABOS algorithm,then,B-Spline surface fitting is done on the grid data.Also,the error of B-Spline is given.Experimental results demonstrate the effectiveness of the presented approaches.