LiveEO has been awarded a high six-digit grant by the German Space Agency at the German Aerospace Center (Deutsche Raumfahrtagentur im DLR), funded by the Federal Ministry of Research, Technology and Space (BMFTR) under the National Programme for Space and Innovation. The funding supports a feasibility study for a new class of artificial intelligence: a Vision Foundation Model (VFM) capable of detecting change across multimodal remote sensing data, combining optical and radar (SAR) imagery in a single, universal architecture.
The 18-month project runs from 1 June 2026 to 30 November 2027 and marks a significant step in LiveEO's effort to move Earth observation AI beyond narrow, task-specific detection toward foundational models built for the full diversity of satellite data.
From Object Detection to a Universal Model
Most AI used in Earth observation today is trained for a single, specific task: identifying a tree, a building, or a change in one type of imagery. These models are powerful but narrow. Each new data source or use case typically requires a new model trained from scratch.
A Vision Foundation Model takes a fundamentally different approach. Rather than learning one task at a time, it learns a general understanding of how the Earth's surface looks and changes across many data types. This makes it possible to adapt a single model to a wide range of monitoring tasks with far less specialized training, and to combine signals that have traditionally been analyzed in isolation.
Why Multimodal Matters: Combining Optical and SAR
Optical and radar (SAR) satellites see the world in fundamentally different ways. Optical imagery captures detail much like the human eye, but depends on daylight and clear skies. SAR penetrates clouds and works day or night, but encodes information that is far harder to interpret. Each modality has long been processed separately, leaving valuable cross-signal insights untapped.
LiveEO's feasibility study investigates whether a Vision Foundation Model can learn to reason across both modalities at once, detecting change reliably even when one data source is incomplete, obscured, or unavailable. For infrastructure operators, this means more consistent monitoring regardless of weather, season, or time of day.
Building Foundational AI for Critical Infrastructure
The project reflects LiveEO's broader strategy: developing AI architecture purpose-built for Earth observation, rather than adapting general-purpose computer vision models to satellite data after the fact. Change detection across optical and SAR is a foundational capability for the company's products, which monitor vegetation, surface movement, and third-party activity along power lines, railways, and pipelines.
"Earth observation is entering a new phase. Until now, we've built specialized models for specialized tasks, but the next frontier is foundational AI that understands satellite data in all its forms. With this project, we're investigating whether a single model can reason across optical and radar imagery the way a skilled analyst would, combining what each sensor sees best," said Dr. Markus Müller, Director of Remote Sensing, SurfaceScout, at LiveEO. "The support from the German Space Agency at DLR and BMFTR allows us to pursue exactly the kind of foundational research that will define the next decade of EO-based monitoring: across infrastructure, the environment, and beyond."
The grant is awarded under the National Programme for Space and Innovation, through which the German Space Agency at DLR supports innovative space technologies developed by German industry on behalf of the BMFTR.
