Combining neural networks for the near-real-time processing of satellite data

This paper presents the combination of neural networks for the operational processing of satellite Earth observation data. Neural networks are used to solve complex and extremely time consuming problems in a novel way. The results obtained are very accurate while the computing time and memory demands are reduced several orders of magnitude compared with traditional solutions. The robustness of the neural networks and the gain in processing speed maintaining an excellent accuracy, allows the near-real-time processing of satellite Earth observation data. Neural networks open new perspectives for the processing of satellite data. They will permit in the near future the on-board processing of Earth observation data for the determination of geophysical parameters in real time.