Application Of ANN In Satellite-Borne Microwave Radiometer

Microwave Sounder is one of the main payloads onboard ChangE-1 Lunar Orbiter. It is a four-channel microwave radiometer which includes 3.0GHz, 7.8GHz, 19.35GHz and 37.0GHz. It is the first time in the history using microwave to detect the moon in orbit.Microwave radiometer is a passive remote sensing apparatus to detect the quantity of radiant energy, and it's a high gain receiver. The output is determined by the system gain and the noise of Microwave Sounder's receiver. Therefore it is very important to lower the amount of fluctuation of the two mentioned above. In 1946, Dicke successfully lower the influence of the system gain thus making the microwave radiometer more practical. In 1967, Gogging completely eliminated the influence of the system gain. In 1968, Hach compensated the fluctuation of the system gain. For the past few years, our country has developed microwave radiometers with high stability, high reliability, high sensitivity, small size and low power consumption. In 1993, Li Jing eliminated not only the influence of the system gain, but also the influence of system noise.First, I calculated the influence on the cold space calibration source by the lunar regolith and the result is up to 10K. Then I calculated the brightness temperature on three specific sites and found that the tendency of the brightness temperature is changed after I took the influence into account. What's more, there are plenty of planets out in the space affecting the cold space calibration source and they're incalculable. So that could definitely affect the calibration.There is a relationship between environmental temperature and system gain, system noise. In order to learn this relationship, we design a temperature experiment to test the performance of microwave radiometer working in the normal temperature range. These results will provide a theoretical reference for the control of microwave radiometer by the neural network.A total power microwave radiometer which is controlled by a neural network is proposed in the end. The system parameters which are obtained in earth-based experiments are kept unchanged in orbit due to the controlling of system gain and offset voltage. This design could dispense with low temperature reference source which is inaccurate and it has the benefits of less physical equipment such as low temperature antenna, lighter weight compared to the traditional periodical calibration at high and low temperature points.