Research On Point Cloud Processing Method Of 3D Laser Scann Technology

With the rapid development of geospatial information service industry,the requirements of geospatial data products are getting higher and higher.Geospatial data is developing towards the direction of large amount of information,high precision,visualization and mining.As a new method of Surveying and mapping,3D laser scanning technology is a technical revolution after the GPS technology in the field of Surveying and mapping.The technology is also called real copy technology,can be established for the region of interest of3 D model detailed in a short period of time,and quantitative analysis,provides a new means to solve problems for the majority of scientific research and engineering and technical personnel.At present,the laser point cloud scanning technology is gradually applied in the field of cultural relic protection,engineering survey,topographic survey,virtual reality,deformation monitoring and traffic accident.However,the study of 3D laser point cloud data collected by 3D laser scanner is still lagging behind.In this paper,we mainly study some problems in the reconstruction of 3D laser point cloud data.In the process of 3D laser point cloud scene reconstruction,different scene data and different purposes make the 3D point cloud data reconstruction method is different.In rendering the scene,researchers pay more attention to the point cloud rendering efficiency;in the cultural relic scene reconstruction process,researchers pay more attention on the part of the reconstruction method of missing data;in deformation monitoring,the researchers pay more attention tothe differences between the different scene description and so on.In this study,multi scene data,single scene scene data,missing data,research on scene reconstruction method of scene data in four cases.In the multi scene data reconstruction,the need for registration of the scene data,the system analysis and synthesis based on the geometric characteristics of the cloud registration method based on improved iterative closest point method based on feature points and presents a method for registration of 3D point cloud data in a 3D scanning system.The method for extracting feature points by SIFT,first get a set of matching points,and then using SVD matrix to calculate the transition parameters R and T decomposition algorithm,and then as the initial value of the iterative closest point method,and the nearest point method and convergence mode are improved,according to the ICP algorithm in Gauss-Newton the approximate value of the slow convergence speed of large error,LM algorithm is used to optimize the projection matrix,LM-ICP algorithm is proposed based on SIFT features.The effectiveness of the proposed algorithm is verified by experiments,and good registration results are obtained.In addition,because of the different proportion of multiple scenes,the existing ICP algorithm can not do multi-scale point cloud registration problem,and a multi-scale scene registration algorithm based on SIFT feature is proposed.In single scene data reconstruction,or when the scene registration data for single scene,the 3D laser scanning point cloud data are usually 3D data,how to reduce the data will influence the efficiency of downsizing and the final reconstruction effect.Three dimensional reconstruction of how quickly the single scene will directly affect the efficiency of the reconstruction of scene data.This study proposes two improved algorithms of point cloud data simplification,a scanning object is based on external attributes(such asdeformation zone and non deformation area,key or non key areas)to classify the 3D point cloud data,non key areas of the bounding box of streamlined,non key regional data is not for data reduction,but this method can not preserve the details of point cloud simplification region.Therefore,the study on the basis of this method we propose another cloud based feature points keep octree partition point cloud simplification algorithm based on octree data point cloud simplification algorithm for non key areas of point cloud data reduction.Finally,the simplified point cloud and the cloud point cloud are used to reconstruct the 3D scene.The two simplified algorithms provide a new solution for the fast reconstruction of 3D point clouds.In the reconstruction of the scene,it is necessary to adopt a reasonable method to extract the change scene data and describe the changing scene data model.This study proposes and ground data points respectively by regional growth method and morphological filtering algorithm to extract the point cloud,the scene data of different changes in the region by using a mathematical model to describe,to describe the changes of the corresponding data in the scene by the angle change between different scene data,in addition,the morphological filtering algorithm improved,which provides a new method for 3D scene change description.In the reconstruction of the missing scene data,due to the lack of instruments or human interference,it is easy to appear the loss of point cloud data,so before modeling,we need to repair the holes in the point cloud.Due to the lack of different parts of the point cloud data point cloud characteristics are different,the manner in which the right of the filling is the key issue.In this paper,the particle swarm optimization SVM model and genetic algorithm to optimize the SVM model for the missing part of the point cloud data to predict,the experimental results show that this method can correctlypredict the point cloud data,which provides a new method for the generation of missing point cloud data.