密集杂波背景下的水下多目标跟踪方法

水下多目标跟踪是水声信号处理领域研究的热点和难点问题。高斯混合概率假设密度(Gaussian mixture probability hypothesis density, GM-PHD)滤波器以其高效的计算效率为解决水下多目标跟踪问题提供了保证。然而，GM-PHD滤波器在跟踪目标时需要先验已知新生目标的强度，否则其性能会出现严重退化。针对该问题，提出一种滑动窗两步初始化高斯混合概率假设密度(sliding window two step initialization GM-PHD, SWTSI-GMPHD)滤波器。将提出的滑动窗两步初始化方法嵌入GM-PHD滤波器，利用滑动窗两步初始化方法估计新生目标强度，减少杂波干扰导致跟踪结果中出现的虚假目标。仿真实验表明，在杂波密集环境下，相较于其他跟踪方法，提出方法将跟踪精度提高69.84%,52.62%和41.05%。

Method of underwater multitarget tracking in dense clutter scenario

Underwater multitarget tracking is a hot and difficult problem in the field of underwater acoustic signal processing. Gaussian mixture probability hypothesis density (GM-PHD) filter provides sufficiently good performance for multiple target tracking problem with its efficient computational efficiency. However, the target birth intensity in GM-PHD filter need to be a priori known when tracking targets, otherwise its performance will decline dramatically. Aiming at this problem, a sliding window two step initialization GM-PHD (SWTSI-GMPHD) filter is proposed. The proposed sliding window two-step initialization method, which can estimate target birth intensity, is integrated into the GM-PHD filter to avoid clutter interference leading to false targets in the tracking results. Simulation results illustrate that the proposed method improves tracking accuracy by 69.84%, 52.62% and 41.05% compared to other tracking methods in a dense clutter scenario.