基于GAUPF-IMM算法的机动目标跟踪研究

摘要

为提高对机动目标滤波跟踪的精度,融合遗传算法与粒子滤波交互多模算法,提出了一种遗传算法优化的交互式多模不敏粒子滤波算法(GAUPF-IMM).该算法采用多模型结构,各模型匹配无迹粒子滤波(UPF),使得新算法用较少的粒子就能体现后验概率密度的特征,减少计算童的同时降低拉子退化现象.并在粒子滤波器输出数据时引入遗传算法对粒子进行重采样,在重要性采样阶段已被优化的基础上进行再次优化,使滤波精度的提高得到双重保障.试验仿真结果验证了新算法的跟踪精度.%In order to achieve a higher accuracy in maneuvering target tracking, a new Interactive Multiple Model (IMM) based on Unscented Particle Filtering (UPF) and Genetic Algorithm (GA) is proposed. The arithmetic adopts multi-model which matches the UPF,and make the characteristics of posterior probability density reflected visibly by less particles. Therefore, the calculation cost is decreased and the particle degeneracy is reduced. The particles are resampled by introducing GA algorithm for the output of the particle filter,thus to implement optimization once more based on the optimization at fundamental sampling, phase, which can guarantee the improvement of filter precision. The experiment and calculation results demonstrate the tracking accuracy of maneuvering target tracking.