Research On Hole Restoration And Simplification Algorithm Of 3D Scattered Point Cloud Data

Three-dimensional Laser Scanning Technology has made great progress in recent years,it has developed rapidly in hardware,but there are also unavoidable problems,For example,in the process of data acquisition,the acquisition of data is often not evenly distributed because of the characteristics of the instrument itself,and in the actual scanning environment,the hole problem may occur frequently due to the occlusion of the line of sight or the improper operation of the operator.What's worse is that the huge amount of data brought by the instrument with high efficiency and precision has brought great trouble to these functions such as displaying,storage,processing and so on.Aiming at the current problem of hole repairing algorithm,which is inefficient or inefficient,we propose a more efficient and effective algorithm.The existing normals based simplification algorithm can largely reduce the point cloud data and preserve the local details of point cloud,but this method is not accurate enough for eigenvalue solution and its efficiency is not high enough.Based on this,this paper improve the solving process of eigenvalue solution,and propose an spatial partition method adaptively.In order to display the results of the algorithm's processing much better,this paper builds a point cloud data processing software based on Qt,OpenGL and PCL.These experimental datas with complex surface features have been selected and the desired results have achieved.The main work of this paper is as follows:(1)The homogenization of point cloud data.In view of the problem of inhomogeneous acquisition of data which collected by 3D laser scanner,on the basis of the existing uniform simplification algorithm,in this paper,a denoising algorithm based on neighborhood analysis is used to denoise the data.Then the minimum outsourced box for the data is solved according to the uniform simplification method.Finally,the point cloud data is divided into voxels according to a certain step length,only the center of gravity of each voxel is retained.(2)A point cloud hole repair algorithm based on mobile least squares is proposed.First,a new method of boundary extraction based on neighborhood point analysis is proposed,and this method is used to extract the hole boundary point.Denoising for the noise points in the boundary point using the denoising algorithm mentioned in the previous article,then the internal and external boundaries of holes are distinguished based on Euclidean clustering,Finally,the moving least square algorithm is used to fill the hole area.(3)An improved simplification algorithm of point cloud normals is proposed.First,we build the kd-tree to find neighborhoods of each point and estimate its normal line for each data point.Second,the angle between the normal of the point and the normal of its K-nearestneighbors points are calculated,According to the related methods,dealing with these K angles,and the result is used as the eigenvalue of the point.According to the eigenvalue,dividing the cloud points into pieces.Different point cloud data may be divided into a different number of members,then the random simplification method is used to simplify these different members.For the lack of flexibility of this method,the related simplification strategy is designed to simplify the point cloud data by any percentage.(4)The construction of point cloud data processing software.In order to display the results of the algorithm's processing much better,this paper builds a point cloud data processing software based on Qt,OpenGL and PCL.This paper uses multiple groups of experiments to carry out the experiment and these results show that these methods of this paper have strong feasibility and robustness.The original point cloud data which is uniform filter the local redundant data much better.The amount of point cloud data is reduced to a certain extent,The point cloud data after repairing holes can almost be consistent with the tested objects and its processing efficiency is higher.The simplification method can simplify most of the data points that are not obvious enough,but it has better effect on feature reservation of point cloud data.The point cloud processing platform is also stable in the actual example.