基矛P—N跟踪器的自适应粒子滤波算法

针对粒子滤波（Particle filter）算法的粒子衰退和计算量过大问题，提出一种将P—N跟踪器与粒子滤波算法结合的目标跟踪方法。首先构造P—N跟踪器，利用跟踪器来确定目标区域范围并输出置信度，以此作为对目标物体定位的依据；在滤波过程中，依据跟踪器结果来进行粒子重采样过程，完成了对抽样粒子集的自适应调节，提高了粒子数量，有效降低了粒子数量。从而达到了抑制粒子衰退和动态调整计算量的目的。实验证明将该方法应用于实时摄像头采集视频跟踪。与传统粒子滤波算法比。在抗粒子衰退与减少粒子数量方面有明显改善。

Serf-adapted particle filter based on P-N tracker

To solve the problem of degeneracy phenomenon and huge computational cost （which decided by the number of the particle）in particle filter, a new solution that combines the P-N tracker with particle filter is proposed. The proposal chooses an appropriate strategy dynamically in the resample step according to the different output of the P-N tracker, such as different range and number of the resample. Thus the degeneracy phenomenon is reduced effectively and the computational is adjusted dynamically. The result of the experiment of on-line tracking used combined mechanism shows the method is better in processing speed and computational than the traditional particle method.