A New Clustering-Based Framework to the Stem Estimation and Growth Fitting of Street Trees From Mobile Laser Scanning Data
Sheng Xu, Xinyu Sun, Jiayan Yun, Hao Wang
Abstract
Estimating individual tree structures from 3-D space may improve the biomass statistics of the urban forest and provide tree-level information for ecological studies. The existing delineation algorithms developed for 3-D point clouds have difficulty in the tree mapping from nonvertical stems or overlapping crowns, and may fail to detect incomplete or occluded branches. Besides, those methods either focus on the individual tree segmentation or crown delineation from the forest, which inadequately estimates the growth fitting of urban street trees. The goal of this article is to present a framework for estimating the growth fitting of street trees' diameter at breast height and under branch height. Tree stems are identified from the achieved street trees' nonphotosynthetic components, including main stems and branches, over different urban trees from mobile laser scanning point clouds. To extract nonphotosynthetic components, a clustering method is proposed to group points from the same stem or branch. The proposed work was validated in both wearable laser scanning data and vehicle laser scanning data, and the experimental scenes contain a range of roadside trees in different structures. In the identification of tree stems, the achieved correctness and completeness are 94.5% and 92.5%, respectively. In the growth fitting, this article calculates a Gaussian model, with the R-square up to 0.81, to describe the growth fitting of Platanus acerifolia. Results show that the proposed approach succeeds in offering applicability over varying street tree types and the improvement for overlapping individual tree information extraction.