The AlignSAR project aims to provide FAIR-guided open datasets and tools designed for SAR applications, ensuring interoperability and consistency with existing and upcoming initiatives and technologies. The project facilitates a wider exploitation of SAR data and its integration and combination with other datasets.
The project aims to achieve the following objectives:
- Define a procedure for creating SAR benchmark datasets for machine learning applications.
- Develop a reference, quality-controlled, documented, open benchmark datasets of SAR spatial and temporal signatures of complex real-world targets with high diversity to serve a wide range of applications with societal relevance. The database will respect FAIR (Findable, Accessible, Interoperable, Reproducible) and Open Science principles.
- Create the database considering both open and closed SAR missions (including at minimum Sentinel-1), maximizing the geographical and temporal coverage, and integrating and aligning multi-SAR images and other geodetic measurements in time and space.
- Define a specification of the signatures and their associated descriptors so that they can be easily indexed, programmatically searched, and retrieved.
- Develop an open-source software library with associated documentation to create, describe, test, validate, and publish SAR signatures, and expand the database.
- Demonstrate, test, and validate the Open SAR Library (database and open-source software) on at least two use cases for machine learning applications.
- Ensure long-term availability of the database and open-source library, potentially through integration with other relevant open platforms and tools.
The project is funded from the European Space Agency (ESA) in response to the ITT ESA AO/1-11394/22/I-DT. It kicked off in February 2023.
Consortium partners: University of Twente, The Netherlands (Lead partner), University of Leeds, United Kingdom, AGH University of Science and Technology, Poland, and RHEA Group, Italy.
