基于三维激光扫描点云数据特征点提取及建筑物重建

发布时间：2018-01-29 01:13

本文关键词： 三维激光点云数据三维建模特征提取　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：点云数据处理与建模是激光扫描系统研究中的一项重要内容。对于不同的目标物体其建模方式也不近相同。常用的建模方式是在网格构建的基础上进行拟合建模,适用于形状复杂,表面变化较大的物体。但网格的构建计算量大,该建模方法常用于数据量较小的物体。对于建筑物而言,数据量大,通常形状较为规则,需要借助专业的建筑物建模软件,可以直接将点云数据导入到建模软件中,然后手动选择点方式来实现模型构建,但是点云数据量大,选点是完全人工的过程,模型构建结果的好坏与建模人员的操作熟练程度直接相关,建模精度得不到保证。而基于特征点的建模方式,是在特征提取的基础上建模,这种建模方法自动化程度得到提高,使建模精度得到一定的保证,并且可以减少人工操作失误造成的精度损失问题。本文阐述了地面三维激光扫描仪的工作原理,对作业中所产生误差分析,讨论减少误差的方法。对点云数据的预处理,包括点云配准、点云去噪、点云精简、数据分割等不同处理方法进行了论述,并对每一种方法的优缺点以及适用性做了探讨。利用Geomagic Studio软件对原始点云数据进行去噪、数据精简等处理。阐释了曲率估计、主成分分析特征提取方法,并利用Matlab编程实现主成分分析法对点云数据特征点提取,利用Imageware软件通过曲率估算对点云数据特征提取,再通过人工手动修改完善特征点提取。最后将点云导入3DMax中,利用手动选点的方式直接对点云数据进行建模,再将点云数据导入Sktechup中,对点云数据特征提取的同时完成建模。最终利用Geomagic Control将两种方式所建模型分别与原始点云数据作对比,生成呈标准分布的标准偏差统计图,偏差值都是集中在一定范围内,说明两种建模方法都是可行的。从3D结果对比表中可看出,基于特征提取建模方式的四种偏差值均比基于手动选点建模方式的小,该实验证明基于特征提取的方法建模能够有效的减少误差,建模精度更好,能有效的提高建模效率。
[Abstract]:Point cloud data processing and modeling is an important part of laser scanning system. The modeling methods for different target objects are not nearly the same. The commonly used modeling methods are fitting modeling based on mesh construction. . It is suitable for objects with complex shapes and large surface changes. However, the modeling method is often used for objects with small amount of data. For buildings, the amount of data is large, but the shape is usually more regular. With the help of professional building modeling software, point cloud data can be directly imported into the modeling software, and then manually select points to achieve model building, but point cloud data is large, point selection is a completely artificial process. The result of modeling is directly related to the proficiency of the modeler, and the precision of modeling is not guaranteed. However, the modeling method based on feature points is based on feature extraction. The degree of automation of this modeling method is improved, and the precision of modeling is guaranteed to a certain extent. And it can reduce the loss of precision caused by manual error. This paper describes the working principle of the 3D laser scanner on the ground, and analyzes the error generated in the operation. The preprocessing of point cloud data, including point cloud registration, point cloud denoising, point cloud reduction, data segmentation and other different processing methods are discussed. The advantages and disadvantages and applicability of each method are discussed. The original point cloud data is de-noised and reduced by Geomagic Studio software, and curvature estimation is explained. Principal component analysis (PCA) is used to extract feature points from point cloud data by Matlab programming. Imageware software is used to extract the feature of point cloud data by curvature estimation, and then manually modify and improve the feature point extraction. Finally, point cloud is imported into 3DMax. The point cloud data is modeled directly by manually selecting points, and then the point cloud data is imported into Sktechup. Finally, we use Geomagic Control to compare the two models with the original point cloud data. The results show that the two modeling methods are feasible, which can be seen from the contrast table of 3D results. The four deviations based on feature extraction are smaller than those based on manual selection. This experiment proves that the method based on feature extraction can effectively reduce the error and the modeling accuracy is better. Can effectively improve the efficiency of modeling.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P225.2

【参考文献】