Automatic Tree Parameter Estimation by Terrestrial Laser Scanner in a Forest Context

Roderik Lindenbergh, Jinhu Wang, Mónica Herrero-Huerta , Beril Sirmacek (2015) "Automatic Tree Parameter Estimation by Terrestrial Laser Scanner in a Forest Context". Technical report (Delft University of Technology)

This report focuses on developing an innovative methodology to automatically estimate the most relevant structural parameters of individual trees sampled by a Terrestrial Laser Scanner (TLS) in a forest scenario. The tree parameters consists of diameter at breast height (DBH), tree location and stem curve. Meanwhile, a Digital Terrain Model (DTM) of the same area is also generated. This study takes the original point clouds as input and first uniformly down-sampling is conducted. Consecutively morphology filtering is employed to segment the points as ground points and tree points. Then for the ground points, interpolation is performed to generated 0.2 meter by 0.2 meter resolution DTM. For the tree points, a voxel based method for tree individualization is used to separate all single trees. DBH is estimated by circle fitting at 20 cm height bin size using RANSAC approach. Tree height is determined by a dimensional analysis from the point cloud after removing noise. The tree location coordinates are obtained by the trunk projection in the principal direction using Principal Component Analysis (PCA). The stem curve is calculated until the canopy begins, through the trunk extraction using voxel down-sampling analysis. The high-quality results evaluated through a sensitive analysis, confirm the feasibility of the proposed methodology. The whole processing runs automatically based on the input several parameters.