融合反射值影像和卡尔曼滤波的三维点云全局拼接方法

发布时间：2018-05-10 06:30

本文选题：三维激光点云 + 全局拼接　；参考：《中国地质大学（北京）》2013年硕士论文

【摘要】：近年来，随着三维激光扫描技术的飞速发展，其在逆向工程、计算机视觉、模式识别、变形监测、曲面检测和摄影测量学等领域有着越来越广阔的发展和应用前景。点云拼接是影响和制约三维激光技术发展与应用的一个重要方面，也是最近几年众多学者和技术人员仍在研究的重点和难点问题。点云拼接作为三维激光扫描技术应用中前期数据处理的一项重要内容，在其应用领域有着非常关键的作用，拼接精度的高低直接或间接决定其应用价值，因此，研究高精度、高可靠性的三维激光点云拼接方法有着重要的意义。本文提出的融合反射值影像和卡尔曼滤波的三维激光点云全局拼接方法是一种高精度、高可靠性的全局动态拼接方法。该方法以地基三维激光扫描系统所采集的上海某段地铁隧道点云为数据源，研究内容主要包括以下几个方面：基于三维激光点云的反射值影像生成，基于反射值影像的特征点提取与识别，基于反射值影像的三维点云拼接，引入控制条件的全局拼接方法，基于卡尔曼滤波和控制条件的三维点云全局拼接。其中，结合地铁隧道的狭长结构特点，采用区段式拼接对传统的相邻站拼接进行改进，使区段内各个测站共用所有的拼接控制点，避免相邻测站之间的误差积累；对于控制信息的引入，考虑到隧道数据结构特点，文中提取隧道中轴线作为拼接控制条件；在隧道中轴线的提取过程中对于断面点的截取，采用改进的BAYSAC算法结合二次参数曲面拟合方法进行局部曲面拟合，并提出了基于模型参数迭代统计的先验概率估计方法。对于文中所提的方法，通过RIEGLVZ系列三维激光扫描仪采集的地铁隧道数据进行了验证，包括反射值影像生成与特征点提取实验、闭合式全局拼接试验、隧道中轴线提取试验以及基于卡尔曼滤波的点云全局拼接实验。其中，采用区段控制式的拼接方法消除和减弱拼接过程中的误差积累，分别以模拟实验数据和实际隧道数据进行了验证；对于隧道中轴线的提取，采用动态的提取方法并引入加权总体最小二乘算法保证提取的中轴线全局一致，保证反馈的卡尔曼滤波信息的精确性。实验结果表明，本文所提出的方法能够快速、稳定的实现大量点云数据的全局拼接，其结果能达到工程应用的精度要求；此外，，三维激光扫描技术在数据的获取和人力要求等方面较传统的数据采集方式都具有明显的优势。
[Abstract]:In recent years, with the rapid development of 3D laser scanning technology, it has more and more extensive development and application prospects in the fields of reverse engineering, computer vision, pattern recognition, deformation monitoring, surface detection and photogrammetry. The point cloud splicing is an important aspect of influencing and restricting the development and application of three-dimensional excitation technology, and is also the latest The point cloud splicing, as an important part of the data processing of the three dimensional laser scanning technology, has a very important role in the application field, and the accuracy of the stitching is directly or indirectly determined its application value. Therefore, high accuracy and high reliability are studied. The method of 3D laser point cloud splicing is of great significance.
The global mosaic method of fusion reflecting value image and Calman filter is a high precision and high reliability global dynamic stitching method. This method is based on the point cloud of a subway tunnel in a section of Shanghai, which is collected by the three-dimensional laser scanning system of the foundation. The research content mainly includes the following aspects: base The reflection value image of the three-dimensional laser point cloud is generated, the feature points are extracted and identified based on the reflection value image, the global stitching method based on the control conditions is introduced based on the 3D point cloud splicing of the reflected value image, and the 3D point cloud based on the Calman filter and the control conditions is integrated. The segmental splicing improves the splicing of the traditional adjacent stations, so that all the stations in the section share all the splicing control points to avoid the accumulation of errors between adjacent stations; for the introduction of the control information, the tunnel axis is extracted from the tunnel data structure, and the extraction process of the central axis of the tunnel is extracted. For the interception of the cross section points, the improved BAYSAC algorithm combined with the two parameter surface fitting method is used to fit the local surface, and a priori probability estimation method based on the iterative statistics of the model parameters is proposed.
For the proposed method, the subway tunnel data collected by RIEGLVZ series 3D laser scanner are verified, including the reflection value image generation and feature point extraction experiment, the closed global stitching experiment, the tunnel axis extraction experiment and the point cloud global stitching experiment based on the Calman filter. The error accumulation in the splicing process is eliminated and weakened, and the experimental data and actual tunnel data are verified respectively. For the extraction of the axis of the tunnel, the dynamic extraction method is used and the weighted overall least square algorithm is introduced to ensure the global consistency of the extracted central axis, and the Calman filtering letter of the feedback is guaranteed. The experimental results show that the proposed method can quickly and steadily achieve a large number of point cloud data, and the results can achieve the precision requirements of the engineering application. In addition, the 3D laser scanning technology has obvious advantages over the traditional data acquisition methods in data acquisition and human requirements.

【学位授予单位】：中国地质大学（北京）
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P225.2

【参考文献】