Journal
LANDSLIDES
Volume 18, Issue 11, Pages 3547-3558Publisher
SPRINGER HEIDELBERG
DOI: 10.1007/s10346-021-01723-4
Keywords
Lidar; Tree matching; Vegetation
Ask authors/readers for more resources
Terrestrial lidar data is useful for monitoring geohazards, such as rockfall and landslides. However, accurately characterizing vegetated landslides with horizontal shear surfaces can be challenging. This paper introduces a novel semi-automated algorithm to extract and calculate the 3D displacement of trees on a slow-moving landslide, improving landslide deformation analysis using point clouds.
Terrestrial lidar data is a powerful resource for monitoring geohazards such as rockfall and landslides. However, vegetated landslides with horizontal shear surfaces remain difficult to characterize accurately due to a lack of exposed and appropriately oriented surfaces with respect to the lidar scanner. As an alternative, the movements of objects on the slope, such as trees, can be used to estimate slope movement. This paper demonstrates a novel semi-automated algorithm to extract and calculate the 3D displacement of trees on a slow-moving landslide, enabling more detailed landslide deformation analysis using point clouds. The method first uses local geometric descriptors to identify raw points corresponding to tree trunks. Several machine learning techniques are compared for this step, and an unsupervised decision tree algorithm with an accuracy of 84% is selected as the final method. After clustering points into individual trunks, trunks are matched through time according to several matching criteria, and their movement and rotation are then calculated using an iterative closest point algorithm. Accuracy of the automated displacement calculations is confirmed through comparison with manual point cloud measurements. A case study is presented demonstrating the method on a landslide through different seasonal vegetation changes, illustrating that algorithm parameters can be quickly adjusted to account for variations in tree species, density of foliage, scanner distance, and other factors. The final matching precision is demonstrated to be between 89 and 99%, indicating a very small number of false-positive matches.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available