基于遗传重采样粒子滤波的弹道跟踪方法

A Ballistic Target Tracking Method Based on Genetic Resampling Particle Filter

在雷达对目标的跟踪过程中,受不同位置回波干扰的影响,测量噪声常表现出以闪烁噪声为代表的非高斯特性,严重破坏了卡尔曼滤波算法的高斯假设,使得卡尔曼滤波器的估计精度大幅下降。同时,不依赖高斯假设的粒子滤波又存在着粒子退化等问题,无法获得良好的跟踪效果。针对这一情况,文中借鉴生物遗传思想,给出了基于遗传重采样的粒子滤波算法,在重采样过程中分别利用粒子的选择、交叉和变异模拟自然界基因的选择、组合和变异,保证了粒子群的全面性和代表性,避免了迭代过程中的粒子退化,解决了测量噪声非高斯条件下的滤波问题。此外,文中将这一方法应用至自由段跟踪中并进行了数学仿真。仿真结果表明:在非高斯噪声条件下,文中所提滤波算法具有较扩展卡尔曼滤波和Huber滤波更好的适应性和更优的估计精度。

In the process of radar tracking to target,the measurement noise always shows non-Gaussian characteristics represented by flicker noise under the influence of echo interference at different positions,which seriously destroys the Gaussian hypothesis of Kalman filter algorithm and makes the estimation accuracy decreased greatly.At the same time,particle filter which doesn′t depend on Gaussian hypothesis has particle degradation problem and can′t achieve good tracking performance.To solve this problem,a genetic resampling particle filter(GRPF)is proposed based on the idea of biological genetics.In the resampling process,the selection,combination and mutation of particles are respectively used to simulate the selection,crossover and mutation of natural genes,which ensures the completeness and representativeness of particle swarm,avoids particle degradation in the iterative process,and solves the filtering problem under the condition of non-Gaussian measurement noise.In addition,the proposed algorithm is applied to free-flight tracking and the mathematical simulation is carried out.Simulation results indicate better adaptability and estimation accuracy of GRPF over the extended Kalman filter and Huber filter under the condition of non-Gaussian noise.