基于改进容积卡尔曼滤波的纯方位目标跟踪

摘要

To solve the nonlinear problem in pure bearing tracking, a range parametrization root-mean-square cubature Kalman filter is proposed to eliminate the effect of the unobservable range and the error introduced by arithmetic operations performed on the finite word-length computer. On the basis of the initial range expressed by several parameterized models, a root-mean-square cubature Kalman filter is run corresponding to every interval, and the probability is computed according to the Bayesian rule. All the weighting outputs of the filters are summed as the final estimation. Simulation results show the root-mean-square cubature Kalman filter can obtain better accuracy and robustness although the complexity of the algorithm is a little too high.%为处理纯方位跟踪的非线性问题,提出了距离参数化均方根容积卡尔曼滤波,在消除距离信息不可测对跟踪影响的同时弱化了计算机有限字长截断效应所引入的误差.在假设目标的初始距离信息用多个参数化模型表示的基础上,对每个模型独立进行均方根容积卡尔曼滤波,并依据贝叶斯准则计算各滤波结果对应的概率,将概率和对应结果的加权融合作为最终滤波结果.实验仿真表明,该滤波虽略微提升了计算复杂度,但获得了更好的滤波精度和鲁棒性.