基于辅助粒子滤波算法的红外目标跟踪

Infrared target tracking base on auxiliary particle filtering algorithm

针对红外目标在跟踪中计算复杂的问题,构建辅助粒子滤波算法。利用贝叶斯重要性采样算法,在权值大的粒子基础上引入辅助粒子变量,然后重新定义重要采样分布函数,防止重采样后粒子概率密度变化。两次加权计算,使粒子权值比仅用重采样的粒子权值变化更稳定,采样点最接近真实状态;同时不同权值粒子的概率阈值可作为粒子滤波是否完成的判断准则。在二维平面构造红外运动目标模型中,系统为零均值高斯白噪声。仿真数据表明:该算法在x,y方向的均方误差、画面处理时间、RM SE性能上优于粒子滤波算法和重采样粒子滤波算法。

In order to solve the problems of complex calculation in the infrared target tracking,the auxiliary particle filtering algorithm was built by the utilization of Bayesian importance sampling algorithm,the introduction of auxiliary particle variables on the basis of large weight particles,and the redefinition of importance sampling distribution function to prevent the change of the particle probability density after re-sampling.The two-weighted calculation makes the change of the particle weight ratio more stable and the sampling point closest to the true state only by the particle weight obtained from the re-sampling,in which the probability threshold of particles at different weight values can be taken as the criterion for judging whether the particle filtering has been completed.In the infrared moving target model structured in the two-dimensional plane,the system is zero-mean Gaussian white noise.Simulation data shows that the algorithm is superior to the particle filtering and resam pling particle filtering algorithms in the mean square error in x and y directions,picture processing,RMSE performance.