基于裂变自举粒子滤波的红外目标跟踪处理

为了研究解决粒子滤波算法的粒子枯竭现象和计算量大的问题,采用了一种裂变自举粒子滤波方法。该方法在测量时对粒子进行裂变自举,其过程为大权值的粒子进行裂变繁殖,裂变繁殖后的粒子数目则正比裂变繁殖前的粒子,然后覆盖粒子群中的小权值的粒子,粒子预平滑处理,同时保持粒子群的特性,再次重抽样中进行粒子防枯竭函数补偿,设置恰当的抽样门限,淘汰权值较低的抽样点,并在保持样本点总数的前提下从权值较高的抽样点中衍生出多个子抽样点,在模型中给出了粒子跟踪均方根误差以及算法步骤,得到了跟踪最佳处理效果。实验仿真用MATLAB语言编程,结果表明,该算法的均方误差为0.36445,优于基本粒子滤波。

Infrared target tracking based on fissile bootstrap particle filters

In order to overcome the particle depletion phenomenon in the particle filter algorithm and the problem of heavy large amount of calculation, a fission bootstrap particle filtering method was used at the moment of measurement. At first, fissile breeding was carried out on the particles with big weights. The particles after fissile breeding are proportional to those before fissile breeding. Then the particles with small weights in the particle group were covered. The particle pre-smoothing was carried out with the characteristics of the particle group maintained. After sampling once again, the function of particle anti-depletion was compensated. An appropriate sampling threshold was set to eliminate the sampling points with smaller weights. Multiple sampling points were derived from the sampling points with bigger weights in the premise of maintaining the total number of sampling points. The root mean square (RMS) error and calculation steps were shown in the model and best tracking effect was obtained. Simulation was carried out with MATLAB. The results show that the RMS error of the algorithm is 0.36445, better than that of the basic particle filtering algorithm.