海量车载激光扫描点云数据的快速可视化方法  

Real-Time Visualizing of Massive Vehicle-Borne Laser Scanning Point Clouds

作　　者：陈驰[1] 王珂 徐文学[1,3] 彭向阳 麦晓明 杨必胜[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]广东电力科学研究院,广东广州510080 [3]国家海洋局第一海洋研究所,山东青岛266061

出　　处：《武汉大学学报（信息科学版）》2015年第9期1163-1168,共6页Geomatics and Information Science of Wuhan University

基　　金：国家973计划资助项目(2012CB725301);海洋公益项目(2013418025);南方电网公司重点科技资助项目(K-GD2013-030)~~

摘　　要：当前车载激光扫描系统的数据量往往达到数十GB乃至TB级,海量激光点云数据的加载与查询对传统可视化方法提出了挑战。本文设计了一种基于内外存调度的三维可视化方法,突破了物理内存对显示数据量的限制。该方法首先利用双层四叉树索引数据结构实现外存储器上的点云数据管理与快速调度,基于该索引动态加载外存储器上的点云数据到内存,从而快速获取海量数据中的实时数据块;然后,利用多线程分时加载双层四叉树索引数据结构,实现激光点云数据外存到内存的实时传输与绘制。实验结果表明,本文方法不受点云数据量与物理内存大小限制,海量点云可视化效果流畅,适用于台式计算机或网络环境下的海量激光点云数据的调度管理与实时可视化。Modern vehicle-borne laser scanning systems are equipped with high sampling rate scanners that can easily collect several billion of points in a single mission. The data volume of the laser data can reach TB grade. The huge data volume puts a strain on to current laser data retrieval and visualization techniques. We designed a quad-tree based data structure for out-of-core management of 3D point clouds, and implemented a dynamic loading scheme for the real time retrieval of 3D point clouds from disks. First, a two level quad tree based data structure is designed for the out-of-core management of 3D point clouds. Dynamic loading scheme is used to achieve real time retrieval of 3D point clouds from disks. A multi-thread visualization component visualizes 3D point clouds in a fluid manner without imposing data size restrictions. Experiments were undertaken to demonstrate the effectiveness of the multi-threaded visualization and low RAM consumption, showing a promising practical solution for the rapid visualization of large data volumes of 3D point clouds that can scale from the local computer to the internet.

关 键 词：车载移动测量 激光点云 四叉树 多线程 细节层次可视化 

分 类 号：P237.3[天文地球—摄影测量与遥感] P208[天文地球—测绘科学与技术]