基于量测噪声方差估计的GM-CBMeMBer算法

GM-CBMe MBer Algorithm Based on Measurement Noise Variance Estimation

针对目标跟踪系统下量测噪声统计特性不准确甚至难以获取的问题,提出一种量测噪声统计特性自适应的高斯混合势均衡多目标多伯努利(Cardinality Balanced Multi-Target Multi-Bernoulli,GM-CBMeMBer)滤波算法。该算法引入Sage-Husa自适应滤波器的思想,利用遗忘因子对量测噪声协方差误差进行修正;建立检验统计量,判断算法敛散性;若滤波发散,则采用有偏估计方法来保证算法收敛性。仿真结果表明在非时变、时变量测噪声方差未知情况下,改进算法的跟踪性能优于传统的GM-CBMeMBer滤波算法,对量测噪声的变化具有较强的适应能力。

In order to solve the problem that the statistical characteristics of measurement noise in target tracking system are inaccurate or even difficult to obtain,Gaussian Mixture Cardinality Balanced Multi-Target Multi-Bernoulli(GM-CBMeMBer)filtering algorithm with Adaptive Noise Statistical Characteristics is proposed.The algorithm introduces the idea of Sage-Husa adaptive filter,uses the forgetting factor to correct the measurement noise covariance error,establishes the test statistic to judge the convergence of the algorithm,and if the filter is divergent,uses the biased estimation method to ensure the algorithm convergence.The simulation results show that the tracking performance of the improved algorithm is better than the traditional GM-CBMeMBer filtering algorithm in the case of unknown time-varying noise variance,which has strong adaptability to the changes of the measurement noise.