Craft Prospect supports successful completion of ESA-funded FOPEN-SAR project to advance long-wavelength SAR processing for concealed object and change detection

Glasgow-based space engineering company Craft Prospect has supported the successful completion of an ESA-funded project exploring how advanced machine learning techniques can unlock the potential of foliage-penetrating synthetic aperture radar (FOPEN-SAR) for future Earth observation applications.

The project was funded by the European Space Agency (ESA) as part of a wider set of activities designed to strengthen European industrial capability through innovative processing technologies supporting Sentinel Expansion missions, including ROSE-L, ESA’s upcoming L-band radar mission.

Led by the University of Edinburgh, FOPEN-SAR focused on the use of long-wavelength spaceborne SAR to detect concealed objects and identify structural changes hidden beneath dense vegetation. These capabilities have significant potential across a range of practical applications, including illegal logging detection, unexploded ordnance monitoring, environmental surveillance, and border security.

Applying machine learning to a challenging new SAR domain

FOPEN-SAR investigated how classical and data-adaptive techniques, which have previously shown promising detection performance in short-wavelength SAR data, can be applied to new long-wavelength P-band and L-band datasets.

This is a technically challenging area. Long-wavelength FOPEN-SAR applications involve complex scattering effects caused by multiple layers of dense vegetation, as well as poorer image resolution and more limited data resources compared to more established SAR use cases.

To help address this, the team developed new training samples and reusable datasets to support further model development, testing, and validation. This included cross-referencing CONAE’s SAOCOM long-wavelength SAR dataset with high-resolution optical imagery from the Airbus Pléiades Neo constellation, alongside the development of change-detection examples and specialist datasets for future work. New machine learning architectures and training strategies were then tailored to the spatial, temporal, and polarimetric characteristics of the data, including self-supervised and fully supervised approaches.

Project partners and Craft Prospect’s role

The technical work was led by the University of Edinburgh, with Dr Mehrdad Yaghoobi acting as Project Manager and Technical Lead, supported by Dr Shuo Li. 

Craft Prospect contributed as a subcontractor, with Lucy Donnell acting as CPL Lead, and Reuben Lynch supporting the work as AI Software Engineer. Craft Prospect’s role focused on use-case specification, validation, impact assessment, and the operational roadmap for future deployment.

Airbus Defence and Space was also involved as a user stakeholder, with Alasdair Helliwell providing expert input and access to data during the project, while ESA’s Dr Thibault Taillade supported the activity through technical guidance.

Project Manager Mehrdad Yaghoobi said of the project:

‘The support of Craft Prospect was crucial towards the end of project when use case selection, and their specific data labelling, needed industrialist inputs. The roadmap for the low TRL achievements was thoughtfully prepared by Lucy Donnell having a great experience on this activity.’

ESA Technical Officer Thibault Taillade stated:

‘The growing availability of globally acquired long-wavelength SAR data from missions such as BIOMASS, ALOS, SAOCOM, NISAR, and the upcoming ROSE-L is reshaping Earth Observation applications by revealing information or changes beneath dense vegetation and enabling the development of operational frameworks for civil-security and resilience applications. This study provides a robust contribution and opens new perspectives for the detection of anthropogenic activity in densely vegetated regions.’

Dr Shuo Li presents FOPEN-SAR at ESA’s INSIGHT 2026 Workshop at ESA-ESRIN in Frascati.

From algorithm development to operational roadmap

Alongside the technical development and testing led by the University of Edinburgh, the project assessed the practical value of FOPEN-SAR capabilities for future mission and user contexts.

The project outputs point toward several important gains for future long-wavelength SAR processing, including:

• Improved insight accessibility, with detected features overlaid on data to support analysis, interpretation, and validation

• Improved insight reliability, through stronger accuracy and robustness scores

• Improved generalisability, with sensor-based feature extraction methods showing potential for application across different mission data

The operational roadmap considers how the processing methodologies developed through FOPEN-SAR could be applied across missions collecting long-wave SAR data. This includes the potential for onboard processing, where there is a valuable opportunity to reduce information latency, compress SAR data volume without losing data value, and deliver the most relevant information to the ground more quickly for action.

For Craft Prospect, this work aligns strongly with the company’s wider focus on intelligent onboard processing, mission autonomy, and low-latency data delivery for next-generation space systems.

Final thoughts

FOPEN-SAR demonstrates how advances in machine learning and signal processing can extend the value of future long-wavelength SAR missions, helping transform complex raw data into more accessible, reliable, and actionable insight for challenging operational environments. 

With greater data access now available than at earlier stages of the activity, there is strong potential to build on the foundations developed by the FOPEN-SAR team.