基于混合粒子滤波的多目标跟踪

Multi-target tracking based on mixtures of particle filtering

针对可变数量的多个红外弱小目标的检测与跟踪问题,提出了基于混合概率密度模型的多目标先跟踪后检测方法,开发了一种t分布混合粒子滤波器。在混合粒子滤波器中,利用每个分量粒子滤波器的输出信息,根据序列似然比假设检验,检测每个被跟踪目标的存在性。通过估计目标在离散占据网格上的出现概率,检测新目标的出现。混合粒子滤波器使用单独的粒子滤波器独立估计每个被跟踪目标的状态,避免算法的计算量随着目标数量增加呈指数增长的问题。仿真实验证明混合粒子滤波器能够跟踪目标数量可变的弱小目标,能够同时检测目标的消失和出现。

For the problem of detecting and tracking a variable number of dim small targets in IR image sequences, a multi-target track-before-detect approach based on the mixture model of probability densities is proposed, and a mixture of t distribution particle filters （MTPF） is developed for the implementation of the proposed approach. In the mixture of particle filters, the existence of each tracked target is detected by using the sequential likelihood ratio test estimated by the output of the component particle filter. New targets are detected by the appearance probabilities in the discrete occupancy grid in the image frame. The proposed approach overcomes the curse of multi-target dimensionality by independently estimating each target state with a separate particle filter, and avoids the exponential increase in the estimation complexity. Importance resampling is carried out in each component particle filter individually. Simulation experiments illustrate that the MTPF algorithm can detect and track the variable number of dim small targets in the IR images. It can also detect the disappearance and appearance of targets simultaneously.