Abstract: In order to promote the target tracking accuracy in noise circumstance, this paper proposes an improved unscented [H]-infinity filter method based on Cauchy decomposition. First of all, Unscented（UT） transform is used to replace the complex Jacobi matrix calculation within the framework of the extended [H]-infinity filtering. The method reduces the approximate truncation errors, and improves the ability to deal with nonlinear system. Second, this method adopts diagonal elements to decompose the root mean square of state covariance matrix. This decomposition method reduces the effect of observation noise variance to filtering estimation precision effectively. The simulation results show that the proposed method obviously improves the object tracking accuracy and stability in different noise statistical properties.

Key words: object tracking, noise statistical properties, unscented transform, Cauchy decomposition, [H]-infinity filtering

摘要： 提升噪声干扰情况下的目标跟踪精度，提出了一种柯西分解改进的无迹[H∞]滤波方法。采用UT变换代替传统的泰勒级数线性近似截断处理，降低近似截断误差，提升处理非线性系统的能力；对目标状态的协方差矩阵均方根进行柯西分解，采用对角元素求解，降低了观测噪声的扰动误差对滤波估计精度的影响。仿真结果表明，所提方法明显提升了不同噪声统计特性情况下的跟踪精度和稳定性。

关键词: 目标跟踪, 噪声统计特性, 无迹变换, 柯西分解, [H&infin, ]滤波