【作者】 杨洋；

【导师】 张永生；

【作者基本信息】 解放军信息工程大学 ， 摄影测量与遥感， 2010， 硕士

【摘要】 车载激光扫描系统的出现为城市三维空间数据采集及建筑物立面模型的构建提供了一种全新的技术手段。它具有快速、精确、直接获取三维信息的特点,在城市三维建模中扮演着日益重要的角色。本文紧紧围绕基于车载LIDAR(Light Detection And Ranging)数据的建筑物立面重建这一主题,重点研究了车载LIDAR数据滤波、分割及建筑物立面特征提取等方面的内容,在理论和算法上取得了一定进展。本文完成的主要工作如下:1.阐述了车载LIDAR系统的组成和对地定位原理,分析了点云数据特点及处理流程,详细介绍了现有的车载LIDAR点云滤波技术、建筑物立面点云分割方法及建筑物立面特征提取技术,归纳和分析了其中需要解决的问题。2.在分析车载LIDAR数据沿扫描线分布空间特征的基础上,提出了一种基于扫描线的车载LIDAR数据滤波方法,并通过实验与基于投影点密度的滤波方法进行了对比分析,证明了该方法的效率和准确性。3.实现了基于随机抽样一致性算法的建筑物立面点云分割。借鉴点云平面方程,引入计算机图形学中的r半径密度设计了相应的判断准则,给出了分割流程。通过分割实验,表明了该方法有利于分割效率的提高。4.阐述了建筑物立面特征自动化识别策略,探讨了墙面和阳台的特征提取方法,提出了一种窗户特征提取的格网化方法,实现了从点云数据中提取出建筑物立面的二维平面特征和三维细节信息,进而自动地重建出具有较高细节层次的建筑物立面几何模型。更多还原

【Abstract】 Vehicle-Borne Laser Scanning System provides a bran-new technical instrument for the three dimensional data collection of city modeling and reconstruction of building facade. It acts a more and more important role in three dimensional city modeling with the character of obtaining three dimensional information at first hand fleetly and accurately. This dissertation focuses on the theme of buiding facade reconstruction based on vehicle-borne LIDAR (Light Detection and Ranging) data closely, and puts keystone on the study of vehicle-borne LIDAR data filtering, segmtation, feature extraction and reconstruction of building facade, which gets along in theory and arithmetic to a certain extent. The major works implemented in this dissertation are listed as follow:1.The components and position principle of vehicle-borne LIDAR system are expatiated. Then the existing methods of vehicle-borne LIDAR point cloud filtering, building facade point cloud segmentation, feature extraction and reconstruction of building facade are analyzed,the issues which need to be settled are summarized.2.A vehicle-borne LIDAR data point cloud filtering method is put forward based on analyzing the distributing space characters of vehicle-borne LIDAR data along scanning beam. The contrast analysis are carried through filtering experiments compared with Density of Projected Points filtering method, and the efficiency and veracity of the method in this dissertation is proved.3.Segmentation of building facade point cloud based on Random Sampling Consensus is implemented. The r-radicus density in Computer Graphics is introduced to devise segmentation judgement rule, and the segmentation flow is presented. Segmentation efficiency is improved, which is indicated in the segmentation experiments.4.Automatic recognizing strategy is expatiated, and feature extraction method is discussed. A method of extracting window features by grids is brought forward. And two dimensional plain character and three dimensional detail information of building facade are extracted from point cloud data, and then building facade geometrical model of high detail levels is constructed automatically.更多还原