Denoising And Detection Of Point Cloud Of Three-dimensional Data

In the production process of industrial tower,in order to allow the tower smoothly assembly site, the tower must be "test assembly" before the appearance of the tower.This process is mainly test tower member right or not,In the processing of tower components,Due to defects in processing technology or operational reasons in the processing,There is a certain deviation in making components and design components.The traditional assembly methods have vertical and horizontal.however,both the assembly are cost,and long cycle,it is difficult to cope with large-scale testing.In this paper, the geometry component product of complex tower quality testing theory and three-dimensional laser scanning technology is applied to detect research.Thereby replacing Industrial Tower members the "test assembly".Methods of obtaining the three-dimensional coordinates.Combined with the needs of the subject,Three-dimensional laser scanner selection and data collection.And through the three-dimensional laser scanner principle inquiry,Analysis of the scan data characteristics,Provide the basis for subsequent data processing.Through the analysis of three-dimensional point cloud noise,The main measurement noise sources, noise analysis of mathematical models,While comparing different Denoising,Here the main comparison Laplace, secondary Laplace, mean curvature flow, bilateral filtering and other methods,Select the appropriate bilateral filtering algorithm in this article,preliminary experiments.For existing bilateral filtering algorithm analysis, after the experiment,Bilateral filtering algorithm has been obtained at the deficiencies of the proposed experiment to improve this article,This article will introduce spatial grid method pretreated point cloud segmentation, and then apply the octree method topological connections, and finally the application field k search feature to search nearby points and achieved the expected results show the improved method proposed by noising effect.At last,tower components quality testing,Ultimate goal is to be scanned point cloud with the design CAD model comparison,Design CAD models need to be discreted,The data of discrete and point cloud scan data coarse registration,According to the proposed based on signatures points rough registration,Then accurate registration based on ICP algorithm,The point cloud models and CAD models to complete registration,Finally, Figure detection method stain deviation described in this article obtained measuring point clouds and CAD models the deviation,and color diagrams and bias data representation.