基于滤波算法的对比研究

The Comparative Research based on the Filtering Algorithms

在处理目标跟踪的过程中,为了与实际接近,动态系统的模型选为非线性,而滤波算法采用非线性系统的滤波方法。在介绍了三种非线性滤波算法(EKF、UKF、PF)的原理和实现的同时,说明了各自适用的范围,以便针对不同问题采取比较便捷的算法来有效地实现算法在实际中的应用。EKF适用于线性化过程中系统对高阶项要求较小的情况下,UKF适用在噪声服从高斯分布的情况下,PF则适用与非高斯分布的情况下。此外,通过实例对三种算法分别进行了跟踪仿真实验,表明UKF、PF算法精度要比EKF算法高,UKF、EKF的实时性比PF好,PF的计算量相对较大。

In dealing with the progress of the target tracking,in order to be close to the practice,the model of the dynamic system is elected to the nonlinear,but the filter algorithms adapt to the nonlinear filtering algorithms. With introducing the principles and the realization of the three nonlinear filtering algorithms,it explains the applicable extension respectively.The different algorithms can be applied in order to settle diverse question.EKF adapt to the situation that the system＇s requirement to the high order items is decreased.The algorithm is the better when the noise is obedient to the distribution of Gaussian.But PF is comfortable to the distribution of non-Gaussian.Through emulation experiment individually to three algorithms in the statistical model of the currency,the results indicate that the UKF,PF have the higher precision than EKF,a character of the better real-time EKF and UKF have is evident,the calculation capacity of the algorithms of PF is larger.