Estimation method of urban green space living vegetation volume based on backpack light detection and ranging

Abstract

Living vegetation volume (LVV) can objectively and accurately reflect the urban greenery quality, and provide a reliable data foundation for the quantitative study aiming to reveal the mechanisms underlying urban greenery ecological functions. According to the characteristics of dispersion and small scale of unit affiliated green space, we proposed a LVV estimation scheme for such urban green space, which included data acquisition, processing, entity segmentation, classification, single tree canopy extraction, and LVV calculation. First, point cloud data was obtained with a backpack LiDAR system, and the ground point clouds were eliminated by a multi-scale algorithm. Second, the Density Based Spatial Clustering of Application with Noise (DBSCAN) algorithm was used to cluster the non-ground point clouds, and density feature-based competitive algorithm was used to re-segmented for the overlapping area to generate independent objects. Third, the PointNet++ network model was used to extracted plant point clouds. Then, the canopy point clouds were extracted using the similarity of principal direction between neighboring points and distribution density of branch and leaf points. Finally, the LVV of individual tree canopy was calculated by the convex hull method, and then the LVV of the accessory greenland was summed up. Taking a science and technology park as an example, its total LVV was 21034.95 m3, among which the number of mango trees was the highest, and the total LVV was the largest (4868.64 m3, accounting for 23.2%). The tree species with the largest LVV per plant was Terminalia neotaliala tree, with an average of 120.37 m3 per plant. The relative error between LVV of trees estimated by this scheme compared with traditional method and convex hull method was 10.7%-33.7% and 2.7%-16.0%, with average value of 20.9% and 8.7%, respectively. This scheme could make full use of the characteristics of the three-dimensional point cloud and use a convex polyhedron to simulate the original form of the tree crown, which was more consistent with the actual situation of trees. The measurement and estimation solution of the LVV provided new ideas for rapid and accurate estimation of urban LVV.