Research On Point Cloud Feature Preservation And Simplification And 3D Modeling Applicatio

The application of point cloud data in 3D modeling is more and more extensive.However,the amount of point cloud data in the original scanning surface is often huge and there is a lot of redundancy.If the original data is directly used for 3D modeling,it will lead to low efficiency and not conducive to the practical application.However,most point clouds simplified by traditional methods have problems such as serious missing of feature points and more holes,and the detailed features of the reconstructed3 D models are not obvious.Aiming at the above problems,this paper studies the following contents:Firstly,introduce the unordered characteristics of point cloud data and its space,and compare and analyze commonly used methods such as neighborhood search,spatial data structure,encoding,etc.for point cloud data,laying the foundation for the calculation of point cloud feature factors in the future.Secondly,a feature preserving point cloud reduction method is proposed to address the problems in traditional methods of simplification.Firstly,feature points are constrained by feature factors such as point cloud normal angle,point cloud curvature,and fast point feature histogram.Then,non feature points are downsampled using grid reduction method.Finally,the extracted feature points and non feature points are fused to form the final reduced point cloud,resulting in prominent feature points with fewer holes and different reduction rates.Through experiments,it has been proven that the proposed method for the Bunny dataset has a reduction error of about 19% compared to the random reduction method and about 86%lower than the curvature reduction method under the same reduction rate.The reduction error of this method for the Armadillo dataset under the same reduction rate is about 38% lower than that of the random reduction method and about 84%lower than that of the curvature reduction method.Next,based on the traditional Poisson reconstruction algorithm,this paper proposes an implicit surface modeling method that simplifies feature constraints and adjusts the consistency of normal vectors.The point cloud surface function is used to solve the vector field of the point cloud and the moving cube algorithm is used to build the isosurface of the model to complete the establishment of the three-dimensional model.Through experiments,it has been proven that the model details reconstructed by this method are prominent.At a 50% reduction rate,the Bunny dataset reduces the number of triangular patches and reconstruction time by about 27% compared to the original data reconstructed model.The Armadillo dataset also reduces the number of triangular patches and reconstruction time by about 25% and 18% compared to the original data reconstructed model,achieving the goal of point cloud reduction.Finally,this article developed a point cloud processing and visualization software that integrates the simplification algorithm and 3D reconstruction algorithm of this article,allowing users to obtain the simplification results and 3D models with only setting parameters.At the same time,some traditional simplification and 3D reconstruction algorithms are also integrated and other point cloud processing functions are added.Finally,the software is tested with different point cloud data to verify the universality of the algorithm in this article.