%0 Generic
%9 Master thesis
%A Weiser, Hannah
%C Heidelberg
%D 2024
%F heidok:35451
%K Vegetation  LiDAR-Simulation
%R 10.11588/heidok.00035451
%T How Does Vegetation Movement During Laser Scanning Affect Common Point Cloud-Derived Metrics? A Virtual Laser Scanning Study
%U https://archiv.ub.uni-heidelberg.de/volltextserver/35451/
%X Close-range, high-resolution laser scanning is used in ecology, forestry and precision agriculture to assess plant health, structure, and productivity, and to monitor growth or regular movement patterns. A common problem in these applications is the wind-induced movement of vegetation during the scanning process, which can lead to multiple representations or blurring of leaves and branches. In this work, we investigate how leaf motion affects common metrics derived from high-resolution terrestrial and UAV-borne laser scanning point clouds of small apple trees. To this end, we use virtual laser scanning of synthetic trees with different levels of animated leaf flutter. Height metrics, voxel metrics, geometric features, and leaf area are calculated from the simulated single-tree laser scanning point clouds and compared across the motion scenarios. Furthermore, the effects of point cloud filtering on terrestrial laser scanning (TLS) data are quantified, both using virtual laser scanning point clouds and a real laser scanning point cloud. We found that leaf motion has significant effects on metrics derived from TLS point clouds, such as the number of leaf and wood points, voxel-based metrics, and geometric features. Leaf area is increasingly overestimated by direct geometric methods as leaf motion increases, with errors up to five times higher than without leaf motion. Meanwhile, standard height metrics and indirect light-extinction based leaf area estimates are fairly robust to leaf flutter. Modelled leaf movement also does not affect the metrics derived from simulated UAV-borne laser scanning point clouds. Leaf area estimation from TLS point clouds using geometric methods can be improved by applying a moving least squares surface approximation filter, which reduces measurement noise and motion effects. This study demonstrates how virtual laser scanning can be used to systematically investigate variables that influence the laser scanning result. Furthermore, by modelling trees as dynamic 3D objects, we take a step towards more realistic simulated point clouds, making virtual laser scanning a powerful method for developing algorithms to analyse 4D vegetation time series.