用于多个机动目标的混合高斯概率假设密度跟踪器

Gaussian-mixture probability-hypothesis-density tracker for multiple maneuvering targets

中文关键词  多目标跟踪  随机集  概率假设密度  混合高斯  机动目标

英文关键词  multiple target tracking  random sets  probability-hypothesis-density  Gaussian mixture  maneuvering targets

中文摘要

      现有的混合高斯概率假设密度(GM--PHD)跟踪器不仅可以估计时变的多目标状态, 还能辨识不同目标并保持其轨迹连续性. 但当多个目标发生机动时, 其稳定性较差, 容易丢失目标. 针对这一问题, 本文提出一种能跟踪多个机动目标的混合高斯概率假设密度跟踪器算法. 算法在GM--PHD滤波的框架上采用修正的输入估计方法将目标的概率假设密度(PHD)表示成混合高斯形式, 并利用不同的标记辨识各个高斯分量, 然后通过PHD滤波方程迭代这些高斯分量和对应的标记, 最终达到跟踪多个机动目标的目的. 仿真实验表明, 和传统的GM--PHD跟踪器相比, 新算法能以更高的稳定性跟踪多个机动目标.

英文摘要

      The existing Gaussian-mixture probability-hypothesis-density(GM--PHD) tracker can estimate time-varying multi-target states, identify different targets and continuously keep on their tracks. However, it is poor in stability and easy to lose targets when multiple targets are in maneuvering. To solve this problem, a Gaussian-mixture probability-hypothesisdensity tracker algorithm for tracking multiple maneuvering targets is proposed. This algorithm employs modified inputestimation technique in the GM--PHD filtering framework to express the probability-hypothesis-density(PHD) of targets in a Gaussian-mixture form, and identifies each Gaussian component by using different labels. The Gaussian components and their labels are iterated through PHD filtering equations, eventually, achieving the tracking of multiple maneuvering targets. Simulation experiment shows that the new algorithm tracks multiple maneuvering targets with higher stability than that of the general GM--PHD tracker.