Estimating Forest Structure from UAV-Mounted LiDAR Point Cloud Using Machine Learning

Monitoring the structure of forest stands is high importance for managers to help them in maintaining ecosystem services. For that purpose, Unmanned Aerial Vehicles (UAVs) open new prospects, especially combination with Light Detection and Ranging (LiDAR) technology. Indeed, shorter distance from Earth’s surface significantly increases point density beneath canopy, thus offering possibilities extraction underlying semantics. example, tree stems can now be captured sufficient detail, which a gateway accurately locating trees directly retrieving metrics—e.g., Diameter at Breast Height (DBH). Current practices usually require numerous site-specific parameters, may preclude their use when applied beyond initial application context. To overcome this shortcoming, machine learning Hierarchical Density-Based Spatial Clustering Application Noise (HDBSCAN) clustering algorithm was further improved implemented segment stems. Afterwards, Principal Component Analysis (PCA) extract stem orientation subsequent DBH estimation. This workflow then validated using LiDAR clouds collected temperate deciduous closed-canopy stand during leaf-on leaf-off seasons, along multiple scanning angle ranges. The results show proposed methodology correctly detect up 82% (with precision 98%) season have Maximum Scanning Angle Range (MSAR) 75 degrees, without having set any parameters segmentation procedure. In future, our method could minimize omission commission errors initially detecting trees, assisting metrics retrieval. Finally, research shows that, under study conditions, within an approximately 1.3-meter height above ground remains low even season, restricting accurate estimation DBH. As result, autonomous UAVs both fly canopy provide clear opportunity achieve purpose.