Research On Feature Preserve Algorithm Of Point Cloud Data

In contemporary society,science and technology develop rapidly.The rate of innovations in all academic disciplines grows every day,so is the application of point cloud data model in 3D solid modeling.The feature keeping of point cloud data has become a hot research topic related to various fields,such as reverse process,3D reconstruction,biomedicine,quality inspection,antiquity protection,and virtual reality,etc.However,the 3D point cloud data obtained in the process of 3D scanning would contain numerous noise points,which do not belong to the original scanning model due to reflections from its surface,human operating factors or ambient lighting changes.The noise in the obtained point cloud data could influence the normal use of subsequent point cloud data and the quality of experimental results.Therefore,in order to obtain satisfactory experimental results,it is necessary to choose an efficient and robust point cloud denoising,feature-preserving algorithm,which is the main purpose of this paper.The traditional feature-preserving algorithms mainly include mesh-based denoising algorithm and normal-based denoising algorithm.Both of them can produce the encouraging results when processing noisy models.However,these feature-preserving algorithms still have the following limitations:(1)Dealing with point cloud models with high noise density,which is difficult to get rid of the noise completely.(2)Because of the sharp features of the model,excessive fairing phenomenon will be occur or the noise will be over-excited,and the holes will appear on the surface of the model.In view of this situation,this paper proposes an improved mesh denoising algorithm and a sharp feature preserving algorithm based on majority voting method,which can achieve feasible and effective noise removal and feature preservation.Firstly,the paper mainly studies the improved mesh-denoising algorithm,this method is mainly by updating the vertex position coordinates to achieve noise removal,and three steps should be conducted as follows: the method firstly pre-processes the original vertices,then filters using the plane normal line,and finally updates the vertex coordinates.The improved algorithm has a strong robustness while preserving point clouds with noise.Secondly,the majority-voting algorithm has been studies in the paper;it is a collective decision-making scheme.The application of the majority-voting feature-preserving algorithm is as follows: firstly,the k-d tree topology of the point cloud is constructed,and the principal component analysis method is used to estimate the normal direction of each point in all neighborhood of the point cloud.Combined with the degree of change of surface,the scattered point cloud is divided into regions.Then the majority voting method is applied to the feature region to identify the fuzzy points.Finally,the threshold is set to remove the determined noise and outliers.The method can avoid problems of incomplete removals or excessive smoothing because of precise points identification.Finally,the corresponding experiments were-conducted to compare the proposed algorithms.The experiment results have proved that the enhanced mesh-denoising algorithm is effective for feature preservation,and have verified the strong robustness of the majority-voting algorithm.