Tree and Plant Modeling

Scalable quantitative parameter extraction and modeling of trees, plants and bushes.

– Background

This project was inspired by the goals of “Peak Carbon Dioxide Emissions by 2030” and “Carbon Neutrality by 2060”, the current lab initialized this project for fine-scale quantitative parameter extraction of trees, plants and bushes through assimilating heterogeneous data sets, such as airborne and terrestrial laser scanning point clouds, hyper-spectral LiDAR and images, etc..

– Goals

The technical and scientific goals of this project are:

1. Leaf reconstruction and fine parameter estimation

Most of the current and existing studies just neglect leaves in biomass estimation. However, the leaves of plants, trees and bushes are of crucial importance in aspects of carbon fixation through photosynthesis, water evaporation by canopy transpiration and wind and noise mitigation. Reconstruction of leaves and fine-scale geometric parameters, i.e. leaf area, leaf aspect and weights, are key elements in subsequent forestry, environmental and ecological applications.

2. Scalable geometric modeling and representation

For trees and plants that have obvious trunks, geometric reconstruction are less challenging than that of bushes. Airborne and terrestrial laser scanning can be utilized for high-precision data collection. Geometric models for different scales of measurements has to be derived for bushes. Efficient data structure and visualization software have to be designed and developed for such applications as well.

3. Biomass estimation and explore subsequent applications

Fine-scale biomass and leaf area estimation are the topmost tasks for the current project. There are quite a number of models for biomass estimation of different types of trees can be employed already. However, fine-scale leaf area needs to use some traditional methods for validation.

– Progress

Data collection

Together with colleagues at Beijing Forestry University, we have collected point clouds and images of a number of outdoor trees, bushes and indoor plants.

Point cloud data of outdoor trees and bushes sampled by terrestrial laser scanning.
Point cloud data of indoor plants sampled by terrestrial laser scanning.
Tree individualization in a forest

The point cloud was collected using an ALS system. The individualization was conducted based on a modified version of my source code project TreeSeparation.

Tree individualization in a forest scenaio. The upper left and upper right images are top and side view respectively. While a cross section view is illustrated by the bottom image.
Leaf modeling

Leaf modeling starts from individual leaf delineation from either point cloud or image data. Currently, we have tested our algorithm on outdoor trees and indoor plants.

Individualization of plant leaves from 3D dense point cloud obtained by terrestrial laser scanning.

The below image is a preliminary result of leaf reconstruction with triangles by Liangliang Nan using Easy3D software. We are still working on the fine-reconstruction of tree leaves.

[Updated until June-30-2022]