EO Point Cloud Generation

This repository aims on the creation Earth Observation Point Cloud. The repository is mainly focus at the implementation of EO-PointCloud as a component into the DeltaTwin service of Destination Earth. Delta Twin allows modeling activities of digital twins and aims to offer a collaborative environment for building and running multi-scale and composable workflows.

This repository contains two implementation of the EO-PointCloud. One using OPEN3D as the core of PointCloud, and the other one is a workaround with plyfile library which is capable of being injected inside the Component of DeltaTwin.

Overview

The project demonstrates how to create a DeltaTwin component and generate a EO-PointCloud File.

1. Access DeltaTwin and see the three mandatory files to create a component:

• Manifest
• Workflow
• Models

In case of doubt and further explanation of how to build a DeltaTwin component, please refers to: DeltaTwin Documentation

2. The folder EO-PointCloud holds the implementation of EO-PointCloud with OPEN3D. For a proper installation of OPEN3D please check the file: install_packages

Quick start

To install the correct environment, it is recomended to use Conda and Conda-Forge. Please follow the steps given at install_packages

Getting Required API keys:

To run any model is necessary having:

• An account at CDSE.
• An account at Destination Earth

CDSE API - Key and Secret

• CDSE key and secret. This is related to fetching the Sentinel-2 images from CDSE ecosystem. Creating an account and fetching data is free for any user! In case of problems finding the key and secret, check it here: FAQ: key and secret

Earth Data Hub

• Earth Data Hub API-key. This is service inside Destination Earth, by having an account with any user has already access to the Earth Data Hub service. Get your API key at: 1 - Log in 2 - Click in your username (upper-right corner) 3 - Account Settings 4 - Add personal access token 5 - Done! Save carefully this token and pass into the Earth_data_hub_api for the main.py

Getting Started

1. Clone this repository

git clone https://github.com/emanuel-gf/EO-Point-Cloud.git

2. Install the required dependencies

conda create --name pointcloud python==3.13.1
conda activate env

Run cli to install all the packages

sh install_packages.sh

3. Run the main.py either at the DeltaTwin folder or EO-PointCloud folder. :

python main.py your_cdse_id your_cdse_access_key your_earth_data_hub_api_key bbox sampled_fraction

DeltaTwin

How to create a DeltaTwin component

For a detailed explanation, see at: DeltaTwin Doc

Deltatwin

1. Install deltatwin

pip install deltatwin

2. Log in

deltatwin login username password -a https://api.deltatwin.destine.eu/

3. Refresh token

deltatwin login

4. Goes to the Folder Deltatwin, fill the inputs_file and start by running locally

deltatwin run start_local -i inputs_file.json --debug 

If you get an error, it may be because the name in the manifest: "name": "eo-point-cloud-generator", should be unique! So you have to create your own name.

5. Push the Component to the Cloud and the Web-UI

deltatwin component publish -t whatever-tag 0.0.1 

EO-POINTCLOUD Component in DeltaTwin

At the DeltaTwin web-UI, it should look like this!

Repository Structure

pcddeltatwin/
├─ DeltaTwin/  ## DELTATWIN COMPONENT OF EO-POINTCLOUD
│  ├─ models/
│  │  ├─ find_bug.ipynb
│  │  └─ main.py
│  ├─ inputs_file.json
│  ├─ manifest.json
│  └─ workflow.yml
├─ EO-Point-Cloud/  ##EO-POINTCLOUD 
│  ├─ src/
│  │  ├─ query_config.yml
│  │  ├─ s3_bucket.py
│  │  ├─ stac.py
│  │  ├─ utils_no2.py
│  │  └─ utils.py
│  ├─ debug_.py
│  ├─ main.py
│  ├─ test_apt_requirements.py
│  └─ tryer.py
├─ initial_setup/
│  └─ install_packages.sh  ## Information on how to properly install OPEN3D and all libraries of EO-PointCloud generation
├─ notebooks/
│  └─ draw_bbox.ipynb ## Quick way to create a bbox and get its boundaries to pass inside the EO-PointCloud arguments
├─ save/
│  ├─ dockerfile  ##Docker with OPEN3D
│  ├─ manifest_bug.json ## Save copy of manifest to workaround with likely bug regarding DeltaTwin and OPEN3D
│  ├─ manifest_test2.json ## Save copy
│  ├─ requirements.txt
│  ├─ t-requirements.txt
├  ├─ copy_manifest.json ## same copy
├  ├─ dockerfile.open3d 
│  └─ test.ipynb  ## Run tests on STAC Catalogue
├─ .gitignore
├─ LICENSE
├─ manifest.json
├─ readme.md
├─ requirements.txt
├─ visualize_pointcloud.py ## Visualize the generated pointcloud

Main.py

You can run any main.py in your environment to test it before trying out the DeltaTwin. Any file main.py requires 5 inputs to run.

main.py -your_cdse_key -your_cdse_secret -your_earth_hub_api_key -bbox -sample_fraction

Utility Functions

The repository includes several utility classes to streamline data processing:

• Sentinel2Reader: For reading and preprocessing Sentinel-2 imagery
• PcdGenerator: For creating point clouds from DEM and imagery data
• PointCloudHandler: For manipulating and saving point clouds

License

Apache License 2.0

Acknowledgments

This dataset is a personal project developed during my internship under Destination Earth initiative within ESA. The majority part of the classes for creation and handling PointClouds were developt by Sebastien Tetauld.

• Sebastien Tetauld
• Copernicus DEM
• Sentinel-2
• DestinE Earth Data Hub