论文部分内容阅读
为解决巡天相机稳像控制精跟踪级系统高精度的光闭环问题,提出一种基于非下采样Contourlet变换(NSCT)去噪预处理和映射最小二乘支持向量机(MLSSVM)回归校正的星点定位方法。针对星图特点,采用自适应的基于NSCT的去噪方法来减小随机误差。从频域角度分析平方质心法系统误差产生的机理,得到其近似解析表达式;利用蒙特卡罗数值仿真的方法,用带有高斯径向基函数(RBF)核的映射MISSVM进行回归分析,得到星点质心的理想位置和系统误差的非线性函数关系,并用它进行系统误差的校正。仿真实验结果表明,提出的方法抗噪能力更强,星点定位精度提高1～2个数量级,具有更为优越的星点定位性能。
In order to solve the problem of high precision closed-loop control of precision tracking-grade system with camera stabilization, a star point based on non-subsampled Contourlet transform (NSCT) denoising preprocessing and MLSSVM regression correction Positioning method. According to the characteristics of the star map, an adaptive NSCT-based denoising method is adopted to reduce the random error. From the perspective of frequency domain, the mechanism of system error of squared centroid method is analyzed and its approximate analytical expression is obtained. The Monte Carlo method is used to conduct the regression analysis with the MISSVM with the kernel of Gaussian radial basis function (RBF) The ideal position of the star’s centroid and the nonlinear function of the system error, and use it to correct the system error. The simulation results show that the proposed method is more robust to noise and improves the positioning accuracy by 1 to 2 orders of magnitude, which results in more superior positioning of the star point.