With the huge amount of data collected by several Earth observation missions, limitations in downlink capacity and storage have become a major obstacle to fully utilising all the valuable information. The CORSA project, funded by the European Space Agency's PhiLab EO Open Science for Society framework, addresses this challenge through a novel AI-based compression method.
CORSA is a family of lightweight AI Compression and Foundation Models for Multi- and Hyperspectral Data. It uses advanced deep learning techniques to efficiently compress imagery from sensors like Sentinel-1, Sentinel-2, and PRISMA, drastically reducing data volume while maintaining high image fidelity, even at 100x compression rates.
CORSA's compressed features can be directly used to build downstream applications such as land-use classification, change detection, and natural disaster mapping. For detailed information about the CORSA method, visit this blog post.
The CORSA compression techniques are now available as openEO processes in the Terrascope backend. With these two new processes, users can compress and decompress Sentinel 2 data efficiently directly into their openEO data processing workflows.
- corsa_compress: Compresses a data cube by means of the CORSA algorithm.
- corsa_decompress: Decompresses a data cube that was compressed by the CORSA algorithm to reconstruct the original data cube.
A working notebook demonstrating these CORSA compression and decompression processes is available in the openEO community examples repository. Users can adapt the notebook to their own applications or use it as a template for integrating CORSA compression into larger workflows.
