测绘学报

• 学术论文 •    

一种八叉树和三维R树集成的激光点云数据高效管理方法

龚俊1,柯胜男2,朱庆3,钟若飞4   

  1. 1. 江西师范大学软件学院
    2. 江西师范大学
    3. 武汉大学 测绘遥感信息工程国家重点实验室
    4. 首都师范大学
  • 收稿日期:2011-04-11 修回日期:2011-12-20 出版日期:2012-08-25 发布日期:2012-08-25
  • 通讯作者: 龚俊

An Efficient Large-scale Point Cloud Data Management Method for Vehicle-borne Laser Scanning Applications

  • Received:2011-04-11 Revised:2011-12-20 Online:2012-08-25 Published:2012-08-25

摘要: 车载激光扫描点云数据已经成为数字城市和危机管理等领域越来越重要的三维空间信息源,针对大规模点云数据高效管理的技术瓶颈,本文提出一种八叉树和三维R树集成的空间索引方法-3DOR树,充分利用八叉树良好的收敛性高效生成R树叶节点,避免逐点插入过程而显著提升索引创建效率,同时R树平衡结构保证数据检索效率。该方法还扩展了R树索引结构支持多细节层次点云模型的生成,并提出了一种多细节层次点云数据组织方法。实验证明,本文提出的方法具有良好空间利用率和空间查询效率,支持多细节层次描述能力和数据缓存机制,为大规模点云数据的后处理与综合应用奠定了坚实基础。

Abstract: Vehicle-borne laser point cloud data has become key 3d spatial information source in fields such as digital city and crisis management. Aiming at technical bottleneck of large-scale point cloud data management, this paper presents a new spatial index method–3DOR-Tree, which integrates Octree and 3D R-tree. This method utilizes Octree to forbid point-by-point insertion and generate leaf nodes of R-tree efficiently. This paper also extends R-tree structure to present levels of detail (LoD) generation algorithm of point cloud models. Finally, a data organization approach is put forward for large-scale point cloud, which easily use file mapping technique to accelerate data access. Experiments prove that this approach has fine space utilization and spatial query efficience with Lod representation capability and data cache mechanism, which lays a solid foundation for post-processing and comprehensive practices of large-scale point cloud data.