【作者】 刘立强；

【导师】 康宝生；

【作者基本信息】 西北大学 ， 计算机软件与理论， 2010， 硕士

【摘要】 随着三维激光扫描技术的发展,人们可以快速准确的获得物体表面大量的采样点。但是这些数据非常庞大,对后续的实时和高效的处理带来了很大的挑战,因此准确且高效的处理这些点云数据,并最终生成逼真的实物模型成为研究的一个重点。本文在此背景之下,对散乱点云数据处理的相关算法进行了研究,其主要研究内容有以下几个方面：1.对空间包围盒分块中栅格边长的计算方法进行了改进。首先以给定边长进行首次划分,计算出划分栅格的黑体占有率；然后综合考虑数据集的范围、点的总数、最近点数k以及黑体占有率的情况下进行第二次划分,使得对不同点云数据的最佳边长计算更为精确。2.在分析了基于点距与基于曲率的精简方法的基础上,综合考虑了数据简化速度与简化精度的前提下给出了一种基于向量夹角的数据简化算法。算法首先计算出采样点及其邻域点的重心点,再计算采样点指向邻域点的向量与采样点指向重心点的向量的夹角,取这些夹角的平均值作为采样点的夹角,依据采样点的夹角值大小来识别特征点；实验结果表明该算法策略能够较好的识别特征点,既提高了简化的速度,又很好的保持了物体表面的几何特征。3.分析了各种已有散乱数据边界点提取算法。在此基础上根据边界点与其邻域点的分布特征,给出了一种基于点距的边界点快速提取算法。算法首先计算采样点及其邻域点的重心点,跟据采样点到重心点的距离与采样到最远邻域点的距离的比值来识别边界点。实验表明该算法能够快速准确的提取三维散乱数据点的边界点。更多还原

【Abstract】 With the development of three-dimensional laser scanning technology, people can quickly and accurately get sample points from the surface of object; but these data points are very large, so it is an enormous challenge to deal with these points real-time and efficiently. Processing the points cloud accurately and efficiently and generating a realistic physical model ultimately will be a research focus. Dissertation studied the scattered point cloud processing and related technologies in this background, and the main research contents are as follow:1. Space bounding box of side length blocking strategy estimation methods are improving. First of all, using a given side length for the first time to divide the grid and calculate the blackbody ratio; and then comprehensive consider the scope of data set, total number of points, k-nearest neighbors points and blackbody ratio for the second division, for a variety of point cloud data making estimate of side length more reasonable, the calculation more efficient.2. Analyzing the methods of data simplification base on point distance and curvature, and then proposes a new algorithm for data simplification base on vector angle over considering the efficient and accuracy of data simplification; Firstly, calculate the center of gravity of sample point and the neighborhood points, and calculate the vectors of sample point to the neighborhood points and the vector of sample point to the center of gravity of points, then calculate the angle between vectors and take the average of these angles as the angle of sample point. According to the angle of the sample point to identify the feature point; The experiment result illustrate that the algorithm can discriminate the feature points and boundary points, keep the geometrical features and boundary points of the surface, and make data simplification more efficient.3. Analyzing methods of data extraction from scattered data, and then proposing an algorithm for data extraction according to the characteristic of boundary points and nearest neighborhood points; Firstly, calculating the center of gravity of nearest neighborhood points, and then identifying the boundary points according to the ratio between the distance of the sample point to the center of gravity point and the distance of the sample point to the farest point of nearest neighborhood points.The experiment result illustrate that the algorithm can extract the boundary points more efficient and exactly.更多还原