多传感器空间配准的病态性及其稳健估计

为提高多传感器空间配准的精度和稳定性,以最小二乘空间配准算法为基础,分析多传感器空间配准病态性产生的原因,指出系统误差估计Fisher信息矩阵的病态性是影响空间配准质量的主要因素。运用条件指标对病态性程度进行评价,并根据信息矩阵的奇异值呈阶梯型分布特点,提出一种混合的奇异值修正稳健估计算法。该算法通过只对较小的条件指标进行修正,实现对系统误差观测方程中不确定成分的有效抑制和确定成分的保留,能较好解决多传感器空间配准系统误差实时估计问题;通过两种不同场景下的系统误差估计结果证明,该算法的使用可显著地提高系统误差参数估计的精度和稳定性。

Ill-condition Analysis and Robust Estimation for Multi-sensor Spatial Registration

In order to improve the accuracy and stability of multi-sensor spatial registration, the causes of the ill condition of the multi sensor spatial registration are analyzed based on LS estimation for multi-sensor spatial registration. The analysis result shows that the ill-condition of the Fisher information matrix is the key factor which influences the quality of system error parameter estimation. The conditional index of information matrix is used to measure the ill-condition of spatial registration. A new algorithm of compound modified singular value decomposition (CSVD) for robust estimation is proposed according to the characteristics of trapezia distribution of the information matrix singular values. The algorithm can effectively suppress uncertainty components and retain deterministic components in the observation equation by modifing only minor conditional index, and can effectively solve the multi sensor real time estimation of spatial registration. Two different scenarios of the estimatied results show that CSVD method can be used to significantly improve the estimation accuracy and stability of system error parameters.