The main objective of our work is to estimate the carbon content of trees in a forest plot. With this in mind, we have implemented the algorithm of Liang et al [1] to detect trees in the wild and measure their diameters at breast height (1.3 m) from point clouds. We usually get our point clouds from photogrammetry techniques, where cameras or direct 3D measurements are used, either with LiDAR or TLS sensors. This document describes the requirements, installation, and how to run our code to detect trees and measure their diameter at chest height.
TreeTool is made up of some sample notebooks and three libraries segTree, TreeTool and Utils. segTree contains several useful functions that allow us to quickly perform operations on point clouds. TreeTool contains our main class. When called, it performs the complete process of tree detection and extraction of diameters at chest height. Finally, Utils contains special functions required by the internal processes and functions for displaying point clouds.
The requirements will depend mainly on the size of the point clouds that will be worked with, since a cloud can vary between hundreds of thousands to hundreds of millions of points. The recommended requirements are designed to process clouds with around ten million points with good fluency. Minimum technical requirements • Windows 10 Processor: 1 GHz or more • RAM: 8 GB • Graphics: DirectX9 or newer
Recommended technical requirements • Windows 10 • Processor: 2 GHz or more, 2 or more cores • RAM: 16 GB • Graphics: Dedicated video card with 4GB of VRAM
Anaconda Python 3.7 packages • pip • open3d • laspy • pdal • python-pdal • gdal • pclpy • matplotlib • pandas • scipy
Anaconda is a package manager system that may make you life easier with your setup. If you do not have it installed in your system, you may want to download it and install it from https://www.anaconda.com/products/individual/get-started
This may be useful to isolate the package installation to your current system setup.
Create a virtual environment, called venv, under the current directory and use python as your interpreter
conda create --name venv python=3.7
To activate the virtual environment run
conda activate venv
Run the following commands to make sure you have the required pieces of software.
- Firstly, install jupyter notebook, a web-based programming environment for python, and other languages!
conda install jupyter
conda install -c conda-forge -c davidcaron pdal python-pdal gdal pclpy laspy pandas
pip install open3d lsq-ellipse jupyter matplotlib scipy
- install Open3d, library for 3d point processing
conda install -c open3d-admin open3d
- Install conda-forge, a collection of recipes, build infrastructure and distributions for the conda manager.
conda config --add channels conda-forge
- Install laspy, a library for reading, writing and modifying files in format LAS for LiDAR data.
conda install laspy
- Install PDAL (Point Data Abstraction Library), and GDAL (Geospatial Data Abstraction Library) for python. The first one will be useful for dealing with point clouds while the second one handles raster and vector data.
conda install -c conda-forge pdal python-pdal gdal
- Install pclpy a python wrapper for the pointcloud library (PCL) developed by davidcaron. It makes it easier to install dependencies for pclpy.
conda install -c conda-forge -c davidcaron pclpy
- Install matplotlib, a python library for plotting static, animated, and interactive visualizations.
conda install matplotlib
- Install pandas, a python library for data manipulation and analysis
conda install pandas
- Install scipy, a python library for scientific computing
conda install -c anaconda scipy
- Install lsq-ellipse a tool for elipse fitting
pip install lsq-ellipse
Finally you can download the point clouds for TLS tests at this address. https://drive.google.com/drive/folders/15aW3Npr9lOdxGrswWrsN9wN0g2Q9pBGo?usp=sharing
The original databases and original publication can be found on this page. https://laserscanning.fi/results-available-for-international-benchmarking-of-terrestrial-laser-scanning-methods/
[1] Liang, X., Litkey, P., Hyyppa, J., Kaartinen, H., Vastaranta, M., & Holopainen, M. (2011). Automatic stem mapping using single-scan terrestrial laser scanning. IEEE Transactions on Geoscience and Remote Sensing, 50 (2), 661-670.