A spatio-temporal deformation model for laser scanning point clouds
A spatio-temporal deformation model for laser scanning point clouds
复制标题
DOI：
10.1007/s00190-020-01352-0
复制
发表时间：
2020-02-11
影响因子：
4.4
通讯作者：
Neuner, Hans
中科院分区：
文献类型：
--
作者：
Harmening, Corinna;Neuner, Hans
地面激光扫描仪的建立改变了工程大地测量的分析策略，从点的方法到面的方法。在最近几年中，关于基于激光扫描仪的几何状态描述的许多发展被做出。然而，区域变形分析仍然是一个挑战。本文将B样条曲面的估计与变形的随机建模相结合，提出了一种时空变形模型。该方法的主要思想是通过三个部分，类似于最小二乘配置：确定性的趋势，代表未失真的对象，随机信号，描述局部均匀变形过程，和测量噪声，占测量过程中所造成的不确定性建模所获得的测量对象。由于以距离相关变差函数的形式对变形进行了随机建模，因此克服了在两个测量时期内定义相同点的挑战。根据由初始趋势面确定的大地基准，可以对所获得的数据集进行点对面和点到点的比较，从而得出可解释和有意义的变形度量。此外，遵循最小二乘配置的基本思想，变形模型允许与时间相关的空间连续描述以及空间和时间连续的变形预测。所开发的方法进行了验证，使用模拟数据集，和各自的结果进行了分析和比较，相对于名义表面。
The establishment of the terrestrial laser scanner changed the analysis strategies in engineering geodesy from point-wise approaches to areal ones. During recent years, a multitude of developments regarding a laser scanner-based geometric state description were made. However, the areal deformation analysis still represents a challenge. In this paper, a spatio-temporal deformation model is developed, combining the estimation of B-spline surfaces with the stochastic modelling of deformations. The approach's main idea is to model the acquired measuring object by means of three parts, similar to a least squares collocation: a deterministic trend, representing the undistorted object, a stochastic signal, describing a locally homogeneous deformation process, and the measuring noise, accounting for uncertainties caused by the measuring process. Due to the stochastic modelling of the deformations in the form of distance-depending variograms, the challenge of defining identical points within two measuring epochs is overcome. Based on the geodetic datum defined by the initial trend surface, a point-to-surface- and a point-to-point-comparison of the acquired data sets is possible, resulting in interpretable and meaningful deformation metrics. Furthermore, following the basic ideas of a least squares collocation, the deformation model allows a time-related space-continuous description as well as a space- and time-continuous prediction of the deformation. The developed approach is validated using simulated data sets, and the respective results are analysed and compared with respect to nominal surfaces.