点云法向量估算研究
发布时间：2019-05-24 20:32
【摘要】：地面三维激光扫描技术作为新兴的测绘数据获取技术,可以快速的获取高密度、高精度反映物体表面信息的点云数据,在逆向工程、文物保护、数字城市、变形监测等多个领域得到广泛应用。获取的点云数据大多数情况下并不具有拓扑信息,需要通过K近邻来建立点之间的空间关系。一致定向的法向量场是许多数据处理工作的基础,而且定向的法向量能够提供下垫面的一阶逼近和内外侧区分。稳健、定向的法向量估算和整个表面的重建工作一样复杂。在数据获取过程中,由于仪器自身存在的系统误差、操作人员水平的高低、外界环境的干扰,使得扫描获取的点云数据中存在噪声与离群点,获取的数据不能真实反映扫描对象的几何信息。目前,主要的点云法向量估算方法是基于主元分析法,主元分析法通过拟合点及其近邻点的局部总体最小二乘平面,用该局部平面的法向量来近似代替点的法向量。算法对噪声有一定的抑制作用,但对离群点较为敏感。需要对点云中含离群点时,如何准确的估算法向量进行进一步研究。论文在点云法向量估算现有研究成果的基础上,针对点云中存在离群点的情况进行研究,主要工作为：1、介绍了地面三维激光扫描技术的发展现状。包括三维激光扫描仪的工作原理、相应的处理软件,介绍了开源C++编程库PCL(Point Cloud Library)及其基本功能,分析了法向量估算和方向调整的重点和难点和研究现状。2、分析了两种类型的K近邻,介绍了对散乱点云数据建立空间索引的必要性,重点阐述和分析了Kd树搜索算法的特点,分析两种点云数据格式,并利用PCL提供的Kd树搜索算法实现了对点云近邻点的搜索。3、介绍了基于Voroni图和局部表面拟合的法向量估算,研究了以局部平面拟合为基础的主元分析法,分析了主元分析法的总体最小二乘本质,推导了主元分析法的数学表达式,并采用主元分析法完成了点云法向量的估算。4、探讨存在离群点时,采用主元分析法估算法向量的误差问题,为了去除离群点,推导了张量投票的闭合解,将点云表示为球张量,采用张量投票算法去除离群点,实验结果表明该算法的有效性。
[Abstract]:As a new surveying and mapping data acquisition technology, ground 3D laser scanning technology can quickly obtain point cloud data with high density and high precision to reflect the surface information of objects, in reverse engineering, cultural relics protection, digital cities. Deformation monitoring and other fields have been widely used. In most cases, the obtained point cloud data does not have topological information, so it is necessary to establish the spatial relationship between points through K nearest neighbors. The uniformly oriented normal vector field is the basis of many data processing work, and the oriented normal vector can provide the first order approximation and the inner and outer side discrimination of the underlying surface. Robust, directional normal vector estimation is as complex as the reconstruction of the whole surface. In the process of data acquisition, due to the systematic error of the instrument itself, the level of operators and the interference of the external environment, there is noise and outliers in the point cloud data obtained by scanning. The obtained data can not truly reflect the geometric information of the scanned object. At present, the main point cloud normal vector estimation method is based on the principal component analysis method. The principal component analysis method uses the normal vector of the local plane to approximate the normal vector of the replacement point by fitting the local total least square plane of the point and its nearest neighbor. The algorithm has a certain suppression effect on noise, but it is sensitive to outliers. It is necessary to further study how to estimate the algorithm vector accurately when there are outliers in the point cloud. On the basis of the existing research results of point cloud vector estimation, this paper studies the existence of outliers in point cloud. The main work is as follows: 1. The development status of ground 3D laser scanning technology is introduced. Including the working principle of 3D laser scanner and the corresponding processing software, this paper introduces the open source C programming library PCL (Point Cloud Library) and its basic functions, and analyzes the key and difficult points and research status of normal vector estimation and direction adjustment. This paper analyzes two types of K nearest neighbors, introduces the necessity of establishing spatial index for scattered point cloud data, focuses on the characteristics of Kd tree search algorithm, and analyzes two kinds of point cloud data formats. The Kd tree search algorithm provided by PCL is used to search the nearest neighbor points of point cloud. 3. The normal vector estimation based on Voroni graph and local surface fitting is introduced, and the principal component analysis method based on local plane fitting is studied. In this paper, the essence of the total least square of principal component analysis is analyzed, the mathematical expression of principal component analysis is deduced, and the estimation of point cloud normal vector is completed by principal component analysis. 4, it is discussed that there are outliers. The principal component analysis (PCA) method is used to estimate the error of the algorithm vector. In order to remove the outliers, the closed solution of Zhang Liang's voting is derived, the point cloud is expressed as a spherical tensor, and the Zhang Liang voting algorithm is used to remove the outliers. The experimental results show the effectiveness of the algorithm.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN249
[Abstract]:As a new surveying and mapping data acquisition technology, ground 3D laser scanning technology can quickly obtain point cloud data with high density and high precision to reflect the surface information of objects, in reverse engineering, cultural relics protection, digital cities. Deformation monitoring and other fields have been widely used. In most cases, the obtained point cloud data does not have topological information, so it is necessary to establish the spatial relationship between points through K nearest neighbors. The uniformly oriented normal vector field is the basis of many data processing work, and the oriented normal vector can provide the first order approximation and the inner and outer side discrimination of the underlying surface. Robust, directional normal vector estimation is as complex as the reconstruction of the whole surface. In the process of data acquisition, due to the systematic error of the instrument itself, the level of operators and the interference of the external environment, there is noise and outliers in the point cloud data obtained by scanning. The obtained data can not truly reflect the geometric information of the scanned object. At present, the main point cloud normal vector estimation method is based on the principal component analysis method. The principal component analysis method uses the normal vector of the local plane to approximate the normal vector of the replacement point by fitting the local total least square plane of the point and its nearest neighbor. The algorithm has a certain suppression effect on noise, but it is sensitive to outliers. It is necessary to further study how to estimate the algorithm vector accurately when there are outliers in the point cloud. On the basis of the existing research results of point cloud vector estimation, this paper studies the existence of outliers in point cloud. The main work is as follows: 1. The development status of ground 3D laser scanning technology is introduced. Including the working principle of 3D laser scanner and the corresponding processing software, this paper introduces the open source C programming library PCL (Point Cloud Library) and its basic functions, and analyzes the key and difficult points and research status of normal vector estimation and direction adjustment. This paper analyzes two types of K nearest neighbors, introduces the necessity of establishing spatial index for scattered point cloud data, focuses on the characteristics of Kd tree search algorithm, and analyzes two kinds of point cloud data formats. The Kd tree search algorithm provided by PCL is used to search the nearest neighbor points of point cloud. 3. The normal vector estimation based on Voroni graph and local surface fitting is introduced, and the principal component analysis method based on local plane fitting is studied. In this paper, the essence of the total least square of principal component analysis is analyzed, the mathematical expression of principal component analysis is deduced, and the estimation of point cloud normal vector is completed by principal component analysis. 4, it is discussed that there are outliers. The principal component analysis (PCA) method is used to estimate the error of the algorithm vector. In order to remove the outliers, the closed solution of Zhang Liang's voting is derived, the point cloud is expressed as a spherical tensor, and the Zhang Liang voting algorithm is used to remove the outliers. The experimental results show the effectiveness of the algorithm.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN249
【参考文献】
中国期刊全文数据库 前10条
1 鲁铁定;周世健;张立亭;官云兰;;基于整体最小二乘的地面激光扫描标靶球定位方法[J];大地测量与地球动力学;2009年04期
2 陈玮娴;袁庆;陈义;;约束总体最小二乘在点云拼接中的应用[J];大地测量与地球动力学;2011年02期
3 郑坤;朱良峰;吴信才;刘修国;李菁;;3D GIS空间索引技术研究[J];地理与地理信息科学;2006年04期
4 丁克良;沈云中;欧吉坤;;整体最小二乘法直线拟合[J];辽宁工程技术大学学报(自然科学版);2010年01期
5 官云兰;詹新武;程效军;周世健;张立亭;;一种稳健的地面激光扫描标靶球定位方法[J];工程勘察;2008年10期
6 陈西江;花向红;鲁铁定;魏成;张言锁;;点云配准中异常同名点探测方法研究[J];大地测量与地球动力学;2013年03期
7 孙同贺;罗志才;;数字化曲线的整体最小二乘平差[J];工程勘察;2013年08期
8 方兴;曾文宪;刘经南;姚宜斌;;三维坐标转换的通用整体最小二乘算法[J];测绘学报;2014年11期
9 李宝;程志全;党岗;金士尧;;三维点云法向量估计综述[J];计算机工程与应用;2010年23期
10 熊邦书,何明一,俞华t，
本文编号：2485161
本文链接：https://www.wllwen.com/kejilunwen/dianzigongchenglunwen/2485161.html
