Research On Multi-resolution 3D Reconstruction Technology Of Karst Caves

The rapid development of network technology and computers has made great changes in our lifestyle.It is often heard that 5G technology,VR technology,cloud computing and other new words.However,many new technologies are not known to be based on three-dimensional stereoscopic visualization.on.Nowadays,underground cave landscape viewing,geological research,development and utilization,and ecological preservation of caves have gradually come into our sight.How to improve the tourism value,aesthetic experience,geological research,and ecological preservation of karst caves will become one of the hotspots in karst cave research.The existence of the karst cave is more meaningful.It requires a comprehensive and multi-level understanding of the karst cave and multidisciplinary integration,such as planning and design,art design,computer science,digital art,environment and natural heritage protection,and geoscience.The solid model becomes the essential basic data.In recent years,new surveying and mapping technologies have been continuously updated,and data collection methods have gradually diversified.Lidar scanning technology has become one of the important equipment for underground 3D data collection.In the early stage,following the project team's participation in the 3D reconstruction of several karst cave projects in Yiliang Jiuxiang Triangular Cave,Wenshanba Meitaoyuan Cave,Wenshanba Meitangna Cave,etc.It was found that usually the three-dimensional reconstruction of karst caves uses a triangle with a uniform size.Network or construct a triangular network based on all point clouds to realize the reconstruction of three-dimensional digital three-dimensional model of karst cave.This type of reconstruction method consumes a lot of time,occupies a large computer memory when processing data,the reconstruction result cannot distinguish between fine parts and rough parts,and the final result data output result file is too large,which requires a higher computer to be used in reconstruction After the results and other issues,this paper proposes the use of multi-resolution 3D reconstruction.Part of the data from the Triangular Cave in Jiuxiang,Yunnan,was selected in the test area.The stalactites in the cave are rich,well-developed,and well-preserved.It is one of the typical underground karst caves in Southeast Yunnan.It has hundreds of large and small karst caves.It is the largest,most numerous,and most complete cave community system in the karst cave landscape with various shapes and colors.It is hailed by experts as "Karst Cave Museum".This article adopts high-precision 3D laser scanning technology to obtain the basic data by performing 3D three-dimensional scanning of the cave.Through data analysis,the algorithm design is combined with the characteristics of the cave itself,and after a large number of experiments,it is compared with more accurate mapping data In the experiment,the point cloud data of karst caves in Jiuxiang,Yunnan were processed,analyzed,and 3D reconstructed.Mainly explore the point cloud data preprocessing method and multi-resolution 3D reconstruction method of karst caves.Including: splicing point cloud data,removing point noise from point cloud data,streamlining useless point cloud data,and in-depth discussion of multi-resolution 3D reconstruction of cloud data,the main research results are as follows:(1)In the data preprocessing process,for the point cloud registration problem,feature points are first used for coarse registration,so that the adjacent point cloud data is integrated,and then the accelerated ICP algorithm is used to repeatedly iterate through point pairs to achieve accurate point cloud.Registration;For the problem of removing noise points from the point cloud,first find out the main reason for generating noisy points,and then use the k-d tree search to find the domain points for denoising.If some noisy points cannot be eliminated,use the form of human-computer interaction In order to reduce the point cloud data,this paper uses the equal density method and the bounding box grid reduction method to reduce the classified point cloud according to the multi-resolution modeling accuracy requirements.For the 3D reconstruction problem,this paper explores the commonly used 3D modeling methods.For example,for regular surfaces,NURBS surface modeling is used.For regular object surfaces,object surface frames are extracted for 3D reconstruction or object surface feature points are extracted for planar modeling.For urban modeling,regional growth triangles are used.The mesh is used to construct a three-dimensional model,and multi-resolution three-dimensional reconstruction is used for complex three-dimensional surfaces.(2)For the three-dimensional modeling method of karst caves,firstly,the improved feature is used to extract the feature values from the geometric features of the nearby points,and the normal vector angle is added as a basis for detecting the feature points.The algorithm implements point cloud segmentation.Experimental results show that the method can realize point cloud segmentation on the 3D point cloud of karst caves,and provide basic data for multi-resolution 3D reconstruction.(3)According to the definition of karst cave characteristic line and the particularity of point cloud of karst cave,according to actual needs,define multi-resolution for karst cave,and carry out 3D reconstruction experiments according to different resolutions.First,the definition of karst cave multi-resolution and accuracy classification are defined,and three-dimensional reconstruction is performed according to the accuracy classification level and different accuracy levels.Compare and analyze the modeling results of this method with ordinary modeling methods from the perspective of modeling performance.Use the data collected by higher-precision equipment to demonstrate the accuracy of this method.Use the relative accuracy value and absolute accuracy value to illustrate the accuracy of this method.Whether the relevant requirements are met.The experimental results show that the method in this paper can achieve efficient and accurate 3D reconstruction of point clouds.Through statistical analysis,the results show that the 3D reconstruction performance and efficiency of the method have been significantly improved,the modeling quality is significantly better than the ordinary modeling method,and has higher practical application value.(4)According to the needs of the application,the three-dimensional digital model should be a whole,easy to use and save.Therefore,a method for stitching and fusion of three-dimensional models with different resolutions is proposed.First,the triangular mesh data is format converted for easy loading into matlab.On the platform;secondly,it traverses the two adjacent stitching mesh boundaries through the triangular mesh spatial geometric relationship to find the boundary to be stitched;then,it uses the constrained Delaunay triangle mesh method to connect and uses B-spline interpolation to make the transition natural;finally,The Laplacian principle is used for smoothing and peak elimination to make the surface smooth.Experiments show that this method can not only be used for splicing fusion of fine models,but also for splicing fusion of large karst caves.Compared with existing software manual splicing,this method has certain automation,can reduce the workload and can guarantee the quality of splicing.In practical engineering applications,it has high practicability.