H.F. de Campos Velho, J.D.S. Silva, E.H. Shiguemori (2007): Hardware Implementation for the Atmospheric Temperature Retrieval from Satellite data, Inverse Problems, Desing and Optimization Symposium IPDO-2007, April 16-18, Miami (FL), USA

Abstract: A new sensor is developed for estimating the vertical atmospheric temperature profiles. The inferred data are obtained by an implementation of an Artificial Neural Network (ANN) on hardware device. Such neurocomputer is configured by using VHDL (Very High speed integrated circuit hardware Description Language). The retrieving methodology is based on an inverse approach in which a Multilayer Perceptron (MLP) network is trained with data provided by a direct model characterized by the Radiative Transfer Equation (RTE). In addition, real radiation data from the HIRS-2 (High Resolution Infrared Radiation Sounder) is used as input for the ANN to generate temperature profiles that are compared to the measured temperature profiles from radiosonde. The generated profiles reveal better estimations, from those results obtained with regularized inversions. The advantages of using ANN based systems are related to the intrinsic features of parallelism, the high retrieving velocity, and the hardware implementation capapability that lead to on-board hardware application in space activities.