论文部分内容阅读
随着遥感卫星在轨任务复杂性的不断提升,对卫星定姿精度的要求也不断提高。星敏感器是星上精度最高的姿态敏感器,因而其在轨标定是提高定姿精度的有效手段。由于大视场星敏感器的镜头畸变复杂,目前广泛采用的基于星对角距的最小二乘法存在一定局限性。因此提出一种基于机器学习的星敏感器在轨标定算法,该方法结合机器学习预测建模思想,通过构造特征建立镜头畸变模型,并结合主成分分析方法进行冗余特征的消除,最后从星角距和模型泛化能力两方面对标定效果进行评价。仿真结果表明:算法对镜头畸变程度较大的星敏感器有良好的校正效果,标定精度始终能保持在0.8″内,与目前几种主流算法相比,具有精度高,鲁棒性好等优点。
With the increasing complexity of on-orbit missions of remote sensing satellites, the requirements on the satellite’s attitude accuracy are also constantly increasing. Star sensor is the most accurate attitude sensor on the star, so its on-orbit calibration is an effective means to improve the attitude accuracy. Due to the complicated lens distortion of the large field star sensor, the least squares method based on the star diagonal distance widely used at present has some limitations. Therefore, a star-sensor calibration algorithm based on machine learning is proposed in this paper. Combining the idea of machine learning prediction modeling, the lens distortion model is constructed by constructing features and the redundancy features are eliminated by principal component analysis. Finally, Angular distance and model generalization ability to evaluate the calibration effect. The simulation results show that the algorithm has a good correction effect on the star sensor with a large degree of lens distortion and the calibration accuracy can always be kept at 0.8 ". Compared with the current mainstream algorithms, the algorithm has the advantages of high precision and good robustness .