Tree Instance Segmentation and Traits Estimation for Forestry Environment Exploiting LiDAR Data Collected by Mobile Robots

Setup

A docker image is provided and details are described in the Docker section below.

Dependencies

This code has been developed and tested on Ubuntu 22.04 LTS using Python 3.10. Currently, we do not support Python 3.8, though it is planned.

System dependencies:

• libegl1 libgl1 libgomp1 (required by Open3D)

sudo apt install libegl1 libgl1 libgomp1

• build-essential (Gives g++ make libc among others)

sudo apt install build-essential

CMake is not needed to build the package. This is handled by the python build system.

forest_inventory_pipeline

After you clone the repository, installing it is a simple

pip install -v .

or

make [install] # install is the default target

Docker

Please check the docker/ folder. A simple Dockerfile is provided that sets up the project in a base Ubuntu 22.04 image. A compose.yaml file is provided for convenience as well. This also has services to run some different experiments on data.

You can run a container with forest_inventory_pipeline setup by running the following from inside docker/

docker compose run fip

Usage

The CLI entrypoint fip covers most of the relevant information needed to use the package. Additionally, parameters are described and their default values are given in corresponding files under config/.

fip --help
fip instance --help
fip dbh --help

Example runs:

# relevant output of the following is an instance_segmented_cloud.ply
fip instance /data/input_cloud.ply --output /output -c config/instance.yaml

In case some stages have been run earlier, corresponding skip flags can be used.

# relevant output of the following is cylinders.pickle and tree_dbh.csv
fip dbh /data/input_cloud.ply --output output -c config/dbh.yaml --skip_gs --skip_norm --skip_cluster

If the --output flag is specified, only then do results get written to disk.

Example Experiments

To test and use the approach, you can use the provided Docker image and some simple experiment scripts in scripts/evaluation. These scripts need the data in a slightly specific format detailed further in their documentation (use --help when running the cli).

Unfortunately, the data used for the paper was collected in context of the DigiForest EU project and is under NDA. This data is not yet released. However, it is planned to soon make a labeled forestry dataset release under the DigiForest project which this repository will also then support. This dataset is expected to be released latest by October 2024. Apologies for the inconvenience.

To run an instance segmentation experiment, cd into the docker/ folder, modify the compose.yaml file to have the appropriate data mount, and then run:

docker compose run fip_instance

The results will get printed to the console.

To run a DBH estimation experiment, cd into the docker/ folder and then run:

docker compose run fip_dbh

Publication

If you use our code in your academic work, please cite the corresponding paper:

@inproceedings{malladi2024icra,
  author = {M.V.R. Malladi and T. Guadagnino and L. Lobefaro and M. Mattamala and H. Griess and J. Schweier and N. Chebrolu and M. Fallon and J. Behley and C. Stachniss},
  title = {{Tree Instance Segmentation and Traits Estimation for Forestry Environments Exploiting LiDAR Data Collected by Mobile Robots}},
  booktitle = icra,
  year = 2024,
  codeurl = {https://github.com/PRBonn/forest_inventory_pipeline}
}

License

This project is free software made available under the MIT license. For details, see the LICENSE file.