Abstract Maneuvering ship tracking has applications in maritime surveillance and the performance of tracking algorithms is affected by background clutter and ship-generated wake. Previous works consider the wake as a form of clutter with some specific spatial distribution. In this paper, the Kelvin wake, which assumes that the ship-generated wake is V-shaped, is modelled as an extended target with an irregular and time-varying shape. Thus, the problem of maneuvering ship tracking in the presence of wake becomes a special multiple target tracking problem, where the ship is a point target and the wake is an extended target. A Gaussian Processes based Multiple Model Joint Probabilistic Data Association (GP-MM-JPDA) method is proposed to estimate the ship state and the wake state simultaneously. Two versions of the GP-MM-JPDA are implemented. The first one can be applied to most multiple target tracking problems and the second one is designed especially for the ship tracking problem. Both versions can improve the ship tracking performance. In addition, the proposed methods can also infer the ship length using the estimated wake angle, which facilitates ship classification. Simulation results show the effectiveness of the proposed methods.

中文翻译：

---

使用高斯过程同时跟踪机动船舶及其尾流

摘要 机动船舶跟踪在海上监视中有应用，跟踪算法的性能受背景杂波和船舶尾流的影响。以前的工作认为尾流是一种具有某些特定空间分布的杂波形式。在本文中，假设船舶产生的尾流为 V 形的开尔文尾流被建模为具有不规则和时变形状的扩展目标。因此，存在尾流的机动船舶跟踪问题成为一个特殊的多目标跟踪问题，其中船舶是点目标，尾流是扩展目标。提出了一种基于高斯过程的多模型联合概率数据关联(GP-MM-JPDA)方法来同时估计船舶状态和尾流状态。实现了 GP-MM-JPDA 的两个版本。第一个可以应用于大多数多目标跟踪问题，第二个是专门为船舶跟踪问题设计的。这两个版本都可以提高船舶跟踪性能。此外，所提出的方法还可以使用估计的尾流角来推断船舶长度，这有利于船舶分类。仿真结果表明了所提出方法的有效性。