Study On Denoising And Feature Extraction Of 3D Point Cloud Data

With the rapid development of 3 d scanning technology,computer technology and information technology,through actual object scanning measurement virtual 3 d model reconstruction of point cloud information technology has become the focus of current research.Its application range is very wide,such as virtual reality model of the building,protection and restoration of cultural relics,the role of 3 d game development and various kinds of mold design and repair and so on.And in the process of model reconstruction,pretreatment of point cloud model(including the denoising smoothing,point cloud to streamline,point cloud segmentation and point cloud splicing,etc.)is a very important part of 3 d reverse modeling.This paper focuses on three dimensional point cloud denoising and pre-treatment of the point cloud point cloud feature extraction algorithm is studied,paper research content is as follows:(1)The estimation of scattered point cloud neighborhood and differential information:In the neighborhood of scattered point cloud,and the calculation theory of differential geometry information research,because of the large amount of scattered point cloud data of this paper,solid chose k-d tree search algorithm topology based on point cloud data.Based on the differential geometry of scattered point cloud information,first of all,according to the theory of curved surface relative contents of point cloud method of vector vector and the curvature of the point cloud is studied,and then put forward based on the sampling point neighborhood local quadric surface fitting method to estimate of the point cloud model differential geometry information.(2)3D point data denoising based on the feature classification:It usually contains noise in the process of obtaining three-dimensional point cloud data,common filtering algorithms have some defects on the model feature-preserving.In order to effectively remove noise and keep the feature information of the model at the same time.In this paper,an improved denoising algorithm was proposed,which based on feature classification of 3D point data.At first,it need to estimate the 3D point cloud differential geometry information by using principal component analysis(PCA)and quadric surface fitting method.And then according to the local feature weights of point cloud mean curvature divide the point cloud data into flat area with less feature information and area with rich feature information,finally according to the different characteristics of the regional using average distance neighborhood filtering algorithm and adaptive bilateral filtering algorithm to filter the noise in point cloud models.By filtering the noise model experiment to verify its feasibility,the results show that the algorithm is able to effectively remove noise at the same time maintained the high frequency feature information of point clouds.(3)Algorithm for Feature Information Extraction on Scattered Point Cloud Data:To solve the problem of efficiency and noise sensitivity in the process of scattered feature extraction.In this pa-per,it proposes a dual-threshold point cloud feature extraction algorithm.First of all,to estimate the differential geometry information of point model through the method of Principal Component Analysis and Local quadratic surface fitting.Then calculate the average normal vector angle in the neighborhood of the sampling point K and the feature weight of average curvature.At last using dual-threshold detection method to extract the feature in-formation of scattered points.Experimental results show that the average time of feature extraction is 0.17s in a noisy environment.The algorithm can extract feature information from scattered point cloud and point cloud model with noise quickly and accurately.Better than the common feature extraction algorithm.(4)Scanning and data processing system building:Using handheld scanner and projection grating scanner for data acquisition,and then based on c + + programming language,using the OpenGL graphics library and VS2013 development tools in three-dimensional point cloud data of denoising and feature extraction algorithm is realized,and by car plaster models for instance that of the whole process of data processing,and feasibility of algorithm.