||
本文为奥地利维也纳技术大学(作者:Bernhard Rainer)的硕士论文,共104页。
本文提出了一种半自动的核外点云形状检测方法。用户控制的交互不是一次对整个点云执行形状检测,而是确定下一步要分割的区域。通过保持尽量小的区域和点的数量,该算法能够在几秒钟内产生有意义的结果。因此,用户会立即收到关于局部几何体的反馈。
由于现代点云可以包含数十亿个点,而且消费类PC的存储容量通常不足以同时将所有点存储在内存中,因此使用一个详细分级的数据结构将点云存储在硬盘上,数据仅在使用时加载到内存中。该数据结构将点云划分为几个小区域,每个区域包含大约5000个点,用于渲染和形状检测。
与点云交互是一项特别艰巨的任务。使用二维lasso相互作用精确选择感兴趣的区域,通常需要多次视图更改和随后的改进。本文仅通过对检测到的形状近似的点集进行选择,提出对lasso相互作用的改进。因此,减少了对前、背景中不希望出现的点的选择。通过使用检测到的形状来改善点拾取,使得仅由该形状近似的点是可拾取的。
本文的研究成果是一个应用程序,它允许用户查看具有数百万个点的点云。它还为快速的局部形状检测和形状辅助交互提供了一种新的交互技术,利用这种局部语义信息来改进用户的工作流。
This thesis presents a semi-automated method for shape detectionin out-of-core point clouds. Rather than performing shape detection on the entirepoint cloud at once, a user-controlled interaction determines the region thatis to be segmented next. By keeping the size of the region and the number ofpoints small, the algorithm produces meaningful results within a fraction of asecond. Thus, the user is presented immediately with feedback on the localgeometry.
As modern point clouds can contain billions of points and the memory capacityof consumer PCs is usually insufficient to hold all points in memory at thesame time, a level-of-detail data structure is used to store the point cloud onthe hard disc, and data is loaded into memory only on use. This data structurepartitions the point cloud into small regions, each containing around 5000points, that are used for rendering and shape detection.
Interacting with point clouds is a particularly demanding task. A preciseselection of a region of interest, using the two-dimensional lasso interaction,often needs multiple view changes and subsequent improvements. This thesisproposes improvements to the lasso interaction, by performing selection only onthe set of points that are approximated by a detected shape. Thus, theselection of undesired points in the fore- and background is reduced. Pointpicking is improved as well by the use of a detected shape, such that only pointsthat are approximated by this shape are pick-able.
The result of this thesis is an application that allows the user to view pointclouds with millions of points. It also provides a novel interaction techniquefor quick local shape detection as well as shape-assisted interactions thatutilize this local semantic information to improve the user’s workflow.
更多精彩文章请关注公众号:
Archiver|手机版|科学网 ( 京ICP备07017567号-12 )
GMT+8, 2024-9-22 18:17
Powered by ScienceNet.cn
Copyright © 2007- 中国科学报社