Individual tree extraction from terrestrial laser scanning data via graph pathing

Individual tree extraction from terrestrial laser scanning (TLS) data is a prerequisite for tree-scale estimations of forest biophysical properties. This task currently undertaken through laborious and time-consuming manual assistance quality control. study presents new fully automatic approach to extract single trees large-area TLS data. data-driven method operates exclusively on point cloud graph by path finding, which makes our computationally efficient universally applicable various types. We demonstrated the proposed two openly available datasets. First, we achieved state-of-the-art performance locating benchmark dataset significantly improving mean accuracy over 10% especially difficult plots. Second, successfully extracted 270 one hectare temperate forest. Quantitative validation resulted in Intersection Union (mIoU) 0.82 crown segmentation, further led relative root square error (RMSE%) 21.2% 23.5% area volume estimations, respectively. Our allows automated access individual level information clouds. The free restricted assumptions It also with an average processing time several seconds million points. expected hoped that would contribute TLS-enabled wide-area qualifications, ranging stand carbon stocks modelling derivation functional traits as part global ecosystem understanding.