激光点云数据压缩及曲面建模研究

发布时间：2018-01-20 03:52

本文关键词： 激光点云数据压缩改进坐标增量法特征保留曲面建模　出处：《西安科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,基于各种数据采集设备生成逼真的三维数字化模型已成为获取物体表面信息的主要手段之一,广泛应用于测量学、计算机视觉、计算机图形学、考古学等领域。现代科技的进步使三维激光扫描硬件设备快速的发展,而相应数据处理技术的发展却相对较慢,成为制约三维激光扫描技术进一步发展的瓶颈,尤其是针对海量点云数据后期的存储、处理、传输和进一步的数据应用。因此,发展激光点云数据处理理论及方法已成为目前学术界研究的热点。本文详细介绍了三维激光扫描技术及点云数据获取与处理的方法,并对点云数据压缩及曲面建模算法进行了深入研究,具体研究工作总结如下:首先,针对三维激光扫描设备采集的点云数据密度大、冗余信息多,现有点云数据压缩算法存在不足的问题,将坐标增量法中一维扫描线逐点压缩扩展到二维扫描线与扫描线间点云数据的压缩,提出了改进坐标增量的点云数据压缩算法,并通过实例,借助Matlab编程,将该算法的压缩效果与坐标增量法、曲率采样法、区域重心法和随机采样法等现有算法的压缩效果比较发现,对于按行列存储的平面或曲面点云数据,该算法具有较好的压缩效果。然后,结合定性与定量的方法从精度、简度、速度及算法通用性等方面对现有算法与改进算法的压缩结果进行了对比分析,验证了改进压缩算法的可行性。并基于Geomagic Studio软件对压缩前后的点云数据进行建模,从表面积变化比和3D标准偏差等方面对本文提出的压缩算法的压缩效果进行了分析,进一步验证了该算法的保真效果与可靠性。最后,针对曲面建模,主要研究了区域生长法、B样条曲面拟合算法、反向传输神经网络法等典型的曲面拟合方法,并总结分析了现有曲面拟合方法中存在的局限性,然后基于Cyclone和Geomagic Studio软件等重建了大雁塔和卢沟桥上石狮子的三维模型,并对重建过程中的不足之处进行了分析。从而为大数据时代下海量点云数据的传输、存储、管理与分析显示提供了一定的参考。
[Abstract]:In recent years, 3D digital models based on various data acquisition devices have become one of the main methods to obtain surface information of objects. They are widely used in surveying, computer vision and computer graphics. Archaeology and other fields. With the development of modern science and technology, 3D laser scanning hardware equipment rapid development, but the development of the corresponding data processing technology is relatively slow, which has become a bottleneck restricting the further development of 3D laser scanning technology. Especially for massive point cloud data storage, processing, transmission and further data applications. Developing the theory and method of laser point cloud data processing has become a hot topic in academic circles. This paper introduces the 3D laser scanning technology and the method of point cloud data acquisition and processing in detail. And the point cloud data compression and surface modeling algorithms are deeply studied. The specific research work is summarized as follows: firstly, the point cloud data collected by 3D laser scanning equipment has a large density and much redundant information. The existing point cloud data compression algorithm is insufficient. The point by point compression of one dimensional scan line in the coordinate increment method is extended to the point cloud data compression between two dimensional scanning lines and scanning lines. A point cloud data compression algorithm with improved coordinate increment is proposed. With the help of Matlab programming, the compression effect of the algorithm is compared with the coordinate increment method and the curvature sampling method. The compression effect of the existing algorithms such as the region barycenter method and the random sampling method is compared. It is found that the algorithm has a better compression effect for the plane or curved surface point cloud data stored by the column and column. Then. Combining qualitative and quantitative methods, the compression results of the existing algorithm and the improved algorithm are compared and analyzed from the aspects of accuracy, simplicity, speed and generality of the algorithm. The feasibility of the improved compression algorithm is verified, and the point cloud data before and after compression are modeled based on Geomagic Studio software. The compression effect of the proposed compression algorithm is analyzed from the surface area variation ratio and 3D standard deviation, and the fidelity effect and reliability of the algorithm are further verified. Finally, the surface modeling is proposed. This paper mainly studies some typical surface fitting methods, such as area growth method, B-spline surface fitting algorithm, reverse transfer neural network method and so on, and summarizes and analyzes the limitations of existing surface fitting methods. Then, based on the software of Cyclone and Geomagic Studio, the 3D model of Lion on the Big Wild Goose Pagoda and Lugou Bridge is reconstructed. The shortcomings of the reconstruction process are analyzed, which provides a certain reference for the transmission, storage, management and display of massive point cloud data in the big data era.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P225.2

【参考文献】