摘要
目标跟踪的模型通常可表示为一个线性的状态方程与一个非线性的观测方程,为提高平方根容积卡尔曼滤波(SCKF)算法的跟踪精度和实时性,提出了一种简化的平方根容积卡尔曼滤波(RSCKF)算法。简化算法在时间更新环节,直接利用状态转移矩阵计算状态变量以及协方差矩阵的一步预测值,避免了原算法中采用一组容积点近似计算的复杂过程,推导证明,简化后的算法其时间更新环节与卡尔曼滤波的一步预测结果一致。最后对两种算法进行了计算复杂度比较以及角跟踪仿真实验。实验结果表明,简化的算法能够降低运算时间并提高跟踪精度。
Considering the fact that the target tracking problems are always modeled by a linear system equation and a non-linear measurement equation, we proposed a Reduced Square-root Cubature Kalman Filter（ RSCKF） to improve the estimation accuracy and the real-time performance. The simplified algorithm utilizes state transition matrix to calculate one-step prediction value of state variable and covariance matrix in time update step, which avoids the complex process of the original algorithm. According to theoretical derivation, the value of time update step in the simplified SCKF is the same as that of the one-step prediction of the Kalman filter. Finally, the two algorithms were used in bearing-only tracking experiment, and the complexity was analyzed quantitatively. Simulation results show that the new algorithm can reduce the operation time effectively and improve the tracking accuracy.
出处
《电光与控制》
北大核心
2015年第3期11-14,共4页
Electronics Optics & Control
基金
国家自然科学基金(61203007)
关键词
目标跟踪
平方根容积卡尔曼滤波
实时性
非线性系统
target tracking
square-root cubature Kalman filter
real time performance
non-linear system
