基于人工神经网络的卫星姿态信号处理

地球遥感卫星的技术难点之一是保持有足够的姿态控制精度．通常卫星的姿态由单一传感器获得，为得到高的姿态精度，需要对多个传感器信息进行综合处理．针对由惯性基准、红外地平仪和太阳敏感器组成的卫星姿态测量系统，提出并设计了带反馈的ＢＰ人工神经网络进行信号处理．研究表明，该方法可以抑制红外地平仪轨道周期误差和卫星进入阴影区时对测量系统的影响，从而使系统的测量精度得以很大提高．

Signal Processing of Satellite Attitude Based on Artificial Neural Network

One of the difficult techniques of a remote sensing satellite is to make its attitude control system to a certain accuracy. Usually, the attitude of a satellite is determined by a single sensor. But in the case that a high accuracy of the attitude is required, we should use several sensors and integrate them in order to process their data. A BP neural network with a feedback loop was proposed and designed for the satellite attitude measurement system which consists of inertial sensors, infrared earth sensors and solar sensors. It is shown that the effects of orbit periodic error of infrared earth sensor and shadow of sun on the accuracy of the measurement system can be restrained. The accuracy of the measurement system is improved a lot.