Abstract: Three-dimensional vegetation structural information is critical for many ecological studies and applications in forestry. It is a non-trivial job to reconstruct accurate and realistic 3D tree model based on the point cloud data collected from terrestrial laser scanners. Although the point cloud data represent the accurate surface information, due to complex structure of trees and occlusion, the data can be noisy and missing. We review the algorithms to extract geometric and topological information from the unorganized point cloud, and procedure for point cloud segmentation, skeleton extraction from point cloud of tree branches and individual 3D tree model reconstruction. We also compare the existing algorithms and suggest the future direction to efficiently and reliably extract information from LiDAR point cloud.

Key words: terrestrial laser scanner, LiDAR, point cloud, tree morphology, tree 3D reconstruction, information extraction