Research On Multiresolution Measurement Point Cloud Integration

With the development of the modern industrial manufacturing,a large number of product parts using irregularly curved surfaces,its design,production,testing and other aspects need to carry out a large number of 3D digital modeling.The 3D digital modeling mainly includes: point cloud acquisition,point cloud registration,point cloud integration,point cloud encapsulation and surface modeling and so on.Among them,point cloud integration is about to integrate multi-view and multi-resolution point cloud into a complete,single-layer,smooth point cloud model,which can be used for later point cloud encapsulation and surface modeling.In this paper,a new multi-resolution measurement point cloud integration method is proposed based on the deep research of existing point cloud integration theories and methods at home and abroad.Compared with the existing point cloud integration methods,the new method has greatly improved the fusion effect and fusion accuracy.(1)Research on point cloud data preprocessing technology.The algorithm for removing noise(clusters)based on incremental clustering is implemented;two vector estimation algorithms are implemented in 2 cases of topological and non topological structure of point data;The methods for calculating the local sampling density and sampling interval of point data are implemented.(2)Research on neighborhood search method for multi-view point cloud,the cylindrical neighborhood search method is implemented.Firstly,the definition of the cylindrical neighborhood is given.Secondly,the search method for the cylindrical data of the point data is given.Finally,the efficiency of the search for the cylindrical neighborhood is validated by the measured data of the point cloud.Compared with the traditional neighborhood,the neighborhood information of the cylindrical neighborhood search is more complete.(3)The relationship between the confidence and the local sampling density of point data is studied: The local sampling density and the confidence of point data are positively correlated.Since the local sampling density of point can be easily calculated according to its 3D coordinates,the local sampling density of the point can be used instead of the confidence as a weighting factor for point cloud fusion.(4)This paper presents a low-density point eating algorithm for surface reconstruction from dense scans.Firstly,the density map for each scan is estimated and the boundary densities are down-weighted.Subsequently,the poorly scanned low-density overlapping points are eaten up based on an user-defined threshold.Finally,the overlapping areas are thinned by using the MLS operator and the homogeneous points are weighted averaged.Based on the above theories,the above algorithms were implemented on the Visual Studio platform using C++ and Open GL open graphics library,and a lot of experimental comparisons were made with existing point cloud integration methods.The experimental results show that the new method has improved both in integration effects and integration accuracy.