机载LiDAR点云数据滤波及建筑物提取技术研究

发布时间：2018-07-10 17:15

本文选题：机载激光雷达（LiDAR） + 点云　；参考：《长安大学》2014年硕士论文

【摘要】：随着卫星导航定位、摄影测量和遥感技术的飞速发展，地理空间信息科学应用领域对获取准确实时可靠的数据要求越来越高。日益兴起的机载激光雷达（Light Detectionand Ranging，LiDAR）技术突破了摄影测量的基本框架，是激光测距、GPS空间跟踪定位、INS姿态确定以及计算机等相融合的一种新兴空间对地观测技术，不受日照和天气等条件的限制，具有全天候、快速、精确等特点，为获取高时空分辨率的地球空间信息提供了极大的便利。机载LiDAR系统有其独到的技术优势，使传统单点定位数据获取变成连续自动数据获取，提高了观测的精度，能够直接获取高精度的飞行区域内地表的三维坐标，可以精确地描述地形的起伏、道路的边缘、植被的树冠和建筑物的复杂构造，在数字地球、智慧地球和数字城市建设等领域有广泛的应用前景。机载LiDAR数据处理中最为关键的步骤是点云数据的滤波和建筑物提取，原始点云数据是一堆杂乱无章的点，需要对点云进行相应处理，从离散的点云中准确地提取地面与地物信息，为实现道路管理、城市规划等提供更有效的信息，因此目前点云数据滤波处理及建筑物提取技术的理论与方法研究已成为国内外众多学者关注的课题。本文以LiDAR点云数据为基础，在无其他辅助数据的情况下，重点探讨了LiDAR点云数据不同格式间的转化、点云数据的三维显示，通过设置不同的插值内插生成DSM深度影像和DSM距离影像，对深度影像数据进行滤波处理，再实现DSM距离影像及滤波后深度影像的建筑物自动化提取，同时通过实验进行具体的验证与比较，从而得出结论。本文的主要工作和研究重点如下： 1.回顾了机载LiDAR系统与数据处理技术的发展研究现状，，介绍了机载LiDAR系统的组成和工作原理，并同传统的航空摄影测量以及InSAR技术工作原理进行比较，同时对LiDAR点云数据的数据特点、误差及处理流程进行详细的分析。本文还对于现存的几种经典数据滤波算法进行原理分析以及对比归纳，简述了自适应滤波器，并进行总结。 2.以ENVI和ArcGIS为主要的操作平台，在ENVI加载用于处理LiDAR数据的插件，实现点云数据不同格式之间的转换，并通过三种不同的表达形式实现LiDAR点云数据的三维化显示。在ArcGIS中导入转换后的点云数据，以反射强度为插值内插生成DSM深度影像，以高程为插值内插生成DSM距离影像，实现对DSM深度影像的滤波处理，用四种不同的自适应滤波进行实验，对结果进行比较和分析。 3.设计一套自动化提取的处理流程，在matlab的平台支撑下，对滤波后的DSM深度影像数据进行二值化操作，并进行先腐蚀后膨胀的形态学开运算，实现部分建筑物的自动化提取，同时对以高程为插值生成的DSM距离影像直接进行二值化及先腐蚀后膨胀的形态学开运算，实现基于DSM距离影像的建筑物提取。最后对两种不同的实验结果和原始点云数据的俯视图进行对比分析，讨论存在的不足并得出结论。
[Abstract]:With the rapid development of satellite navigation, photogrammetry and remote sensing technology, the application of geospatial information science is becoming more and more demanding for obtaining accurate and reliable data. The rising airborne laser radar (Light Detectionand Ranging, LiDAR) technology has broken through the basic frame of the perturbation measurement, which is the laser range finding, the GPS space heel Tracking location, INS attitude determination and computer and other integration of a new space to earth observation technology, not limited by sunshine and weather conditions, all weather, fast, accurate and so on, provide great convenience for obtaining high spatial and temporal spatial information of the earth space. The airborne LiDAR system has its unique technical advantages, making the tradition Single point location data acquisition becomes continuous automatic data acquisition, improves the accuracy of observation, and can directly obtain high precision three-dimensional coordinates of the surface of the flight area. It can accurately describe the undulating terrain, the edge of the road, the crown of the vegetation and the complex structure of the building, which is led by the digital earth, the intelligent earth and the digital city construction. The domain has a wide range of applications.
The most important step in the airborne LiDAR data processing is the filtering of the point cloud data and the building extraction. The original point cloud data is a heap of disordered points. It is necessary to deal with the point cloud and extract the ground and ground information from the discrete point cloud, so as to provide more effective information for the realization of road management and urban planning. The theory and method of filtering processing of pre cloud data and the theory and method of building extraction technology have become a subject of attention of many scholars at home and abroad. This paper, based on LiDAR point cloud data, focuses on the transformation of different formats of LiDAR point cloud data in the absence of other auxiliary data, and the three dimensional display of point cloud data, by setting different data. Interpolation is interpolated to generate DSM depth image and DSM distance image, filter the depth image data, and then realize the automatic extraction of the building of the DSM distance image and the filtered depth image. At the same time, the concrete verification and comparison are carried out through the experiment, thus the conclusion is obtained.
The main work and research focus of this paper are as follows:
1. review the development and research status of airborne LiDAR system and data processing technology, introduce the composition and working principle of the airborne LiDAR system, compare with the traditional aerial photogrammetry and the principle of InSAR technology, and analyze the data characteristics, error and processing flow of the LiDAR point cloud data in detail. Several classical data filtering algorithms are analyzed and compared. The adaptive filters are summarized and summarized.
2. take ENVI and ArcGIS as the main operating platform, load the plug-in for processing LiDAR data in ENVI, realize the transformation between different format of point cloud data, and realize the three dimensional display of LiDAR point cloud data through three different expressions. The converted point cloud data is introduced in ArcGIS, and the reflection intensity is interpolated to create DSM depth with the reflection intensity. In degree image, DSM distance image is generated by interpolation interpolation, and the filtering processing of DSM depth image is realized. Experiments are carried out with four different adaptive filters, and the results are compared and analyzed.
3. a set of automated extraction process is designed. Under the support of the platform of MATLAB, the filtered DSM depth image data is operated on two values, and the morphological opening operation is carried out after corrosion and expansion. The automatic extraction of some buildings is realized. At the same time, the two values of the DSM distance image generated by the elevation are first valued and first. The morphological opening operation of the expansion after corrosion is carried out to realize the building extraction based on DSM distance image. Finally, the two different experimental results and the original point cloud data are compared and analyzed, and the shortcomings are discussed and the conclusions are drawn.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;TN958.98

【参考文献】