三维激光扫描点云精简研究

发布时间：2018-06-26 04:28

本文选题：点云精简 + 包围盒法　；参考：《东华理工大学》2015年硕士论文

【摘要】：随着三维激光扫描技术的发展,硬件和软件更新换代,使用成本下降,越来越多的领域因此获得了极大的便利。然而随着精度要求不断提高,点云数据量大,噪声点多等问题日益成为三维激光扫描技术发展的瓶颈。如何快速准确地去除噪声点,精简冗余点云成为研究的热门领域。论文在总结三维激光扫描技术的现状和采集工作流程的基础上,以点云数据精简为主要研究对象,分析了原始数据到符合用户要求过程中的重难点,本文的研究内容和成果如下所示:1.在众多精简算法和模型中,选择了针对点云数据集合本身的四种算法:基于包围盒均匀精简法,基于包围盒k邻域二次曲面拟合法,基于八叉树k邻域二次曲面拟合法和基于八叉树k邻域法向夹角法。通过定性定量比较,对其异同点,精确度、简化度和效率性等指标进行分析,直观的表现出各种精简算法的适用性和优劣性,并提出新问题:(1)对现有方法进行改良,对阈值的选取进行阶梯分化,(2)如何搜索最近点形成k邻域,或者如何规避k邻域的使用。2.根据对比现有精简算法,对两个问题进行了探索,分别提出了解决方法,并从其思路来源,原理,理论模型方面预测了其可行性,且在具体算例中进行了精简演算。此外,还通过控制变量对其进行了与已知方案的横向对比和自身的纵向对比,得到可行化研究结论,展望未来发展趋势,提出见解。
[Abstract]:With the development of 3D laser scanning technology, the hardware and software are updated and the cost is reduced. More and more fields have been greatly facilitated. However, with the continuous improvement of precision, the problems of large amount of point cloud data and many noise points have become the bottleneck of the development of 3D laser scanning technology. How to quickly and accurately remove noise points and reduce redundant point clouds has become a hot research area. On the basis of summarizing the current situation of 3D laser scanning technology and collecting work flow, this paper takes point cloud data reduction as the main research object, and analyzes the heavy and difficult points in the process from raw data to meeting the user's requirements. The contents and results of this paper are as follows: 1. Among many reduced algorithms and models, four algorithms are selected for point cloud data set: uniform reduction method based on bounding box, Quadric surface fitting method based on k-neighborhood of bounding box, Based on octree k-neighborhood Quadric surface fitting method and octree k-neighborhood normal inclusion method. Through qualitative and quantitative comparison, the similarities and differences, accuracy, simplification and efficiency of the index are analyzed, which directly show the applicability and advantages of various simplified algorithms, and put forward new problems: (1) to improve the existing methods, The threshold is divided into steps. (2) how to search the nearest point to form k neighborhood, or how to avoid the use of k neighborhood. According to the comparison of the existing reduction algorithms, the two problems are explored, and the solutions are put forward respectively, and the feasibility is predicted from the sources, principles and theoretical models of the methods, and the simplified calculation is carried out in a concrete example. In addition, the control variables are compared with the known scheme and their own longitudinal comparison, and the feasible research conclusions are obtained, the future development trend is prospected, and some opinions are put forward.
【学位授予单位】：东华理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TN249

【参考文献】