Multi-sensor architectures for thermal imaging missions

Among the innovative offerings that NewSpace companies bring to the space industry, one of them is constellations of thermal instruments.

Thermal infrared (TIR) imagers are inherently more challenging to calibrate and geolocate than visible and near-infrared (VNIR) instruments due to:
• their lower spatial resolution
• higher sensor noise levels and,
• scarcity of stable reference features.
This makes radiometric and geometric modelling more complex. Multi-sensor satellite architectures that co-locate VNIR and TIR imagers are increasingly valuable for Earth observation applications. These applications require accurate, timely surface temperature data. Integrating reflective and emissive sensors on a single platform unlocks opportunities to improve geometric precision and radiometric consistency.

Our goal with this research is to evaluate whether multi-sensor architectures offer an opportunity to enhance the geolocation accuracy of lower-resolution TIR sensors when combined with higher-resolution VNIR imagers through cross-calibration and shared spatial models. We explore how thermal imager data streamlines processing workflows for bottom-of-atmosphere products, enabling rapid and reliable insights. Multi-sensor solutions have their advantages, but also pose challenges, such as time-synchronisation between imagers and onboard systems. Acquiring reference data with common features across spectral bands for spatial calibration remains difficult.

We compare two fundamentally different processing strategies for VNIR-TIR multi-sensor satellites:
• a purely reference-based workflow
versus
• a cross-imager approach that relies on reference data for VNIR sensors alone

Our findings will state the strengths and limitations of each method. We will present results from multiple real-world space missions, with different instruments onboard. These results will inform the design of future thermal imaging constellations.