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． One of the technical difficulties of Earth’s remote sensing satellites is to maintain sufficient attitude control accuracy. Usually the satellite attitude obtained by a single sensor, in order to obtain high attitude accuracy, the need for a comprehensive treatment of multiple sensor information. Aiming at the satellite attitude measurement system composed of inertial reference, infrared horizon and sun sensor, a BP artificial neural network with feedback is proposed and designed for signal processing. The research shows that this method can restrain the impact of the orbit period error of the infrared horizon and the satellite’s influence on the measurement system when entering the shadow area, so that the measurement accuracy of the system can be greatly improved.