This repository contains CGI's open-source GeoData360 processing services for Project Seagrass.
The Project Seagrass services are designed to support identification, monitoring, and measurement of seagrass in UK coastal waters.
This repository currently describes a supervised classification workflow, including pre-processing of input data, training of a machine learning model given some ground truth data, and use of the model to produce a map of seagrass.
The services are intended to be run as workflows on the CGI GeoData360 platform, but individual steps can be developed and tested independently.
These services are still in development, and are regularly evolving. The steps are implemented with a variety of tools, including gdal and Orfeo Toolbox. New tools, patterns, and implementations are welcomed for integration.
The training and classification steps require standardised inputs. These currently include:
- Satellite imagery
- Sentinel-2 L2A products (sen2cor is invoked if required)
- Provided by the Copernicus programme
- Reference: https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-2
- Bathymetry data
- GEBCO_2022 Grid
- Reference: https://www.gebco.net/data_and_products/gridded_bathymetry_data/
- Seagrass habitat suitability model
- Provided by Project Seagrass and Swansea University
- Reference: https://www.frontiersin.org/articles/10.3389/fmars.2022.997831/full