Research On Extraction Of Boundary Points From Planar Point Cloud

With the development of modern computer technology and manufacturing technique,more and more important surveying technology,such as the airborne LiDAR,terrestrial laser scanning,vehicle mobile laser scanning and so on,emerges in the traditional surveying field.The speed of traditional methods to acquire data is slow,the amount of data is less and it is hard to obtain accurate complete data of the scanning target.3D laser scanning technology is able to obtain the superficial point data and images data of scanning target,overcoming the defects above in surveying.This paper mainly studies problems on data organization methods,neighbor query,point cloud registration and target setting,boundary point extraction from planar points.The main contents are as following:(1)Experiment about the speed of data read,data organization and neighbor query,point cloud visualization of Octree and kd-tree are conducted based on the third point cloud library.The experiment results show that kd-tree is superior to Octree in data organization and neighbor query.Furthermore,the multi-processors can accelerate the speed of algorithm.(2)Registration between stations is complex and the targets setting influence the accuracy of registration in the preprocess stage.Taking these problems into account,a simple model for registration is proposed.After comparing several models of registration,the results show that the model can simplify existing complicated process,realize gross errors process with 3 common targets and improve the stability of calculating result.What's more,the simple model can replace seven-parameter model in small region.At last the paper gives principle of targets setting in practical data collection to ensure data quality after studying the influence of target setting on the accuracy registration.(3)Multi-threshold method to boundary points from planar point cloud is proposed based on existing methods to extract planar boundary points.And the setting principle of threshold is analyzed theoretically and experimentally.The experiment results of comparing different threshold,show that the proposed algorithm can accurately and fast extract boundary points compared with the blocks,vector composition and angles variance algorithms.Especially the boundary points of point cloud with inner hole can be accurately extracted by combining the principle of setting threshold(4)Aiming at the problem of grid size in alpha shapes algorithm,the formula to calculate the size of grid size is given,which combines the density of point cloud.After abundant calculation and experiment,the results show that he ratio of the size of grid and density of point cloud changes between 1.0 and 1.8,the time consuming is the least and the extracted boundary points are the same with the situation without using topological relation to extract...