The goal of the LC-CMS use case is to perform fully automated, efficient, and repeatable global Land Cover (LC) mapping for small and mid-scale features. It is a generic GEM use case which shows global Continuous Monitoring capabilities of the GEM project. LC-CMS use case is of special interest to TomTom as it provides a more efficient, scalable, and maintainable alternative to the previous TomTom LC production pipeline (Earth Cover Engine) based on the capabilities of eo-learn.
To achieve this objective, an end-to-end machine learning based pipeline was built, capable of detecting Level 2 land classes of GEM Taxonomy using mid and low-resolution Sentinel-2 spectral bands. The pipeline was developed using eo-grow and eo-learn frameworks (GEM processing framework) developed by Sinergise. eo-grow in particular allows to easily scale-up and distribute the workload.
We demonstrate the results obtained over Africa in a lightweight web demonstrator app: