为了提高跟踪性能，许多跟踪算法都加入了目标回波幅度信息。采用幅度特征的局限性之一是目标的信噪比（SNR），即，通常假定振幅似然的参数是已知的和恒定的。在实际应用中，目标信噪比总是未知的，并且依赖于方位角，因此它会波动。本文提出了一种混合标记多贝努利（LMB）滤波器，它将信号幅度引入LMB滤波器，用于跟踪未知和波动信噪比的目标。目标信噪比的波动由自回归伽马过程建模，并考虑了Swerling 1和3目标的幅度似然。在Rao-Blackwell分解下，提出了一种基于Laplace变换和Markov Chain Monte Carlo方法的近似Gamma估计器来估计目标信噪比，并在目标信噪比的条件下，采用高斯混合滤波器来估计运动状态。通过一个包含三个交叉目标的跟踪场景分析了该混合滤波器的性能。仿真结果验证了所提出的SNR估计器的有效性，并量化了将幅度信息用于多目标跟踪的好处。
英文摘要：
The amplitude information of target returns has been incorporated into many tracking algorithms for performance improvements. One of the limitations of employing amplitude feature is that the signal-to-noise ratio (SNR) of the target, i.e., the parameter of amplitude likelihood, is usually assumed to be known and constant. In practice, the target SNR is always unknown, and is dependent on aspect angle hence it will fluctuate. In this paper we propose a hybrid labeled multi-Bernoulli (LMB) filter that introduces the signal amplitude into the LMB filter for tracking targets with unknown and fluctuating SNR. The fluctuation of target SNR is modeled by an autoregressive gamma process and amplitude likelihoods for Swerling 1 and 3 targets are considered. Under Rao-Blackwell decomposition, an approximate Gamma estimator based on Laplace transform and Markov Chain Monte Carlo method is proposed to estimate the target SNR, and the kinematic state is estimated by a Gaussian mixture filter conditioned on the target SNR. The performance of the proposed hybrid filter is analyzed via a tracking scenario including three crossing targets. Simulation results verify the efficacy of the proposed SNR estimator and quantify the benefits of incorporating amplitude information for multi-target tracking.
