基于迭代卡尔曼粒子滤波的目标跟踪算法研究

粒子滤波在基于图像序列的目标跟踪中获得了广泛应用．针对其计算量较大的问题，提出一种迭代卡尔曼粒子滤波算法，将非线性跟踪问题分解为线性子结构的全局状态空间模型和非线性子结构的局部状态空间模型，利用粒子滤波在卡尔曼滤波估计值的局部范围内搜索目标，逼近真实目标状态．将实验结果与粒子滤波进行比较，结果表明，迭代卡尔曼粒子滤波减少了粒子数，降低了计算量，能够对高机动目标进行实时稳定的跟踪．

Iterative Kalman-Particle Filter for Object Tracking

Particle filters（PF） are widely applied for various visual tracking problems but restricted with computation load,and then hierarchical Kalman-particle filter （IKPF） based on coarse to fine strategy is proposed to reduce the computation for real-time tracking. The algorithm regards the significant nonlinear system as the combination of a linear state space describing global motion and a nonlinear state space describing local motion. As the optimal linear filter,Kalman filter（KF） is introduced to predict the global motion state. Then Particle filter is used to search the true ob-ject state in the local area based on the estimation by Kalman filter. The comparisons between IK-PF and PF are undertaken, and the results indicate that IKPF need less particles and reduce the computation, and can achieve successful object tracking in real video sequences in which the target objects undergo rapid and abrupt motion.