UC4: Forestry and Biodiversity

This repository contains Forestry and Biodiversity monitoring and analysis algorithms created for Sentinel-2 satellite and UAV mounted hyperspectral camera data



Report Bug - Request Feature

Table Of Contents

• Summary
• Structure
• Models
• Authors
• Acknowledgements

Summary

Within this repository, you'll discover various models and computational tools designed for forestry and biodiversity. Models created for tree health monitoring system from Sentinel-2 multispectral data, tree health and fire monitoring using UAV mounted hyperspectral cameras, and automatic wild boar monitoring and detection from thermal cameras.

Structure

The repository folders are structured as follow:

• data: here you should add the Boar, Hyperspectral and Sentinel-2 datasets that you download from Zenodo.
• models: models developed for forestry and biodiversity monitoring
• libs: libraries created for hyperspectral and sentiinel-2 data processing
  • hyperspectral-lib: Python3 library created for hyperspectral data analysis, reading, writing and modelling purposes.
  • sentinel-tools-lib: Python3 library created for sentinel data processing and analysis, with the capabilities of downloading sentinel data from AWS with the help of sentinelhub-py library.
• platform.json: Structured information about the models and their parameters.
• README.md: This file, providing an overview of the repository.

Models

The models developed are the following:

Wild boar monitoring and detection model created using YOLOv5

This model was created using YOLOv5 object detection framework with and extension algorithms created to count unique wild boars in gathered UAV thermal data.

Sentinel-2 data preprocessing algorithms and pipeline

Algorithms created for automatic raw Sentinel-2 multispectral data processing to a usable dataset based on regions of interest.

Automated region of interest monitoring models based on Sentinel-2 data

Models created to automatically monitor regions of interest using Sentinel-2 data and extends the Sentinel-2 preprocessing algorithms.

Drone data processing pipelines and algorithms

Algorithms created to process raw hyperspectral and thermal data gathered by the UAVs into datasets usable in modelling.

Forest health monitoring model using Sentinel-2 multispectral data

Model created for monitoring forest health using processed Sentinel-2 multispectral and additional data.

Hyperspectral data processing algorithms and Hyperspectral calibration plate detection model

Base hyperspectral data processing algorithms and model created to automatically detect calibration plates in hyperspectral mission that are used to calibrate hyperspectral data from radiance values to reflectance.

Automated hyperspectral data clustering model with Convolutional Autoencoders

Model created to automatically find data clusters (data classes) in hyperspectral data cubes for further analysis base on Convolutional Autoencoders.

Forest health monitoring model using hyperspectral data

Model for forest health monitoring on processend and clustered hyperspectral data.

Forest fuel evaluation model using hyperspectral data

Model for forest fuel evaluation on processend and clustered hyperspectral data.

Authors

Vytautas Paura - ART21 - Vytautas Paura

Acknowledgements

This project is funded by the European Union, grant ID 101060643.