Developing an eScience technology platform
Microwave imaging, in particular Synthetic Aperture Radar, both airborne and space-borne, has a growing number of applications in Earth Observation, such as oceanography, marine climatology, ice monitoring, environmental monitoring, tropical forest monitoring (deforestation), many forms of deformation monitoring for geology, volcanology, land subsidence, disaster monitoring and warning (earth quakes, land slides, floods, oil spills), and infrastructure, and also in planetary exploration, i.e. mapping of planet surfaces (e.g. Mars, Venus).
Furthermore SAR has important applications in Security and Defense. As microwave sensors become increasingly capable of high resolution and multi-channel data acquisition, real-time processing of sensor data is becoming a limiting factor in the advance of real-time monitoring applications, either because of the sheer amount of computation, or because of the size and weight constraints of the sensor platform. Approximating and computationally efficient SAR imaging algorithms have been used for many years, because of these same limitations, but these limit the scope and ultimate accuracy of the final imaging result.
Helicopter equipped with TNO’s AMBER SAR system
AMBER is a unique SAR system in that it combines energy efficient FMCW radar technology with digital beam forming techniques, allowing novel SAR modes performing far outside the boundaries of existing conventional small SAR systems. AMBER produces a continuous raw sensor data flow of 7 Gb/s, requiring ~TFLOPs computation for image formation The system capabilities are mainly limited by practical limitation of the processor. Digital beam forming SAR is also the technology projected for the next generations of earth observation radar satellites.
This project will focus on designing a real-time processing framework for AMBER’s SAR data. Specifically, we will focus on three aspects: (1) design and implementation of HPC algorithms (parallel and accelerated) for SAR processing, (2) modeling the performance of SAR processing algorithms to enable adaptive resolution-to-performance tradeoffs, and (3) build a proof-of-concept framework by instantiating the architecture with the parameters matching the performance requirements of AMBER.