顾及自适应多细节层次的八叉树点云管理算法

为了解决大规模点云不易有效组织、动态可视化时冗余度大,且较难实现自适应显示的问题,提出顾及细节层次(levels of detail, LOD)的八叉树点云管理算法.该算法基于八叉树索引将扫描点限定在每个结点范围内,利用自上而下空间分割和自下而上参数计算相结合的预处理策略,减少实时阶段计算量,通过构建保守性模拟误差,使场景各处均可自动满足可视要求,并辅之以高效加速方法,实现了点云的有效组织和自适应流畅显示.实验研究表明,在优化的预处理和辅助加速策略支持下,与经典R树算法相比,该算法实时阶段计算量小,每帧自适应漫游平均时间在0.04 s以内. 

Management Algorithm of Point-Cloud Data Based on Octree Concerned with Adaptive Levels of Detail

Large-scale point-cloud data are not easy to organize effectively and have great redundancy at dynamic visualization, and it is hard to realize the adaptive display. Aiming at these problems, a new algorithm concerned with the levels of detail (LOD) of point-cloud expression on the basis of octree structure was proposed. The algorithm assigned every scanning point into an octree node, and integrated top-down division with down-top calculation as the pretreatment strategy to reduce the amount of real-time calculation. Then it made any region meet the accuracy requirement and display speed automatically by building conservative simulation-error evaluation criteria. Furthermore, with the help of acceleration methods, large-scale point-cloud data could be organized effectively and expressed smoothly with little data redundancy. Preliminary experiments show that the algorithm has abilities to overcome the shortcoming of the classical R-tree methods; meanwhile, with the support of optimized pretreatment and assistant acceleration methods, the amount of real-time calculation is small and the time of each frame can hold within 0.04 s easily.