基于改进遗传粒子滤波与SME的多目标跟踪算法

摘要

To investigate the non-Gaussian problem caused by symmetry transformation while tracking multiple targets with symmetric measurement equation, which could not be solved by traditional filtering methods,an improved genetic particle filter is proposed. The advanced method uses the negative correlation between the noise component and the weight of particles to improve the probability density function upon which the weight calculation is dependent in the stage of update,and avoids the calculation of the new measurement noise. Meanwhile,genetic algorithm can help increase the use efficiency and diversity of particles as well as avoid the filter divergence and local optimization. Simulation was made,and it turned out that the filtering performance of advanced Genetic Particle Filter is better than Extended Kalman Filter,Unscented Kalman Filter and Joint Probability Data Association filter in multi-target tracking based on symmetric measurement equation.%针对基于对称量测方程的多目标跟踪，传统的滤波手段无法解决因对称变换带来的非高斯问题，提出一种新的遗传粒子滤波方法。新的滤波算法利用粒子的噪声含量与权值的负相关，改进了更新过程中权值计算所依赖的概率密度函数，避免了新量测噪声的求解。同时利用遗传算法的优势，保障了粒子的多样性，提高了粒子的使用效率，防止了滤波发散及局部最优。仿真结果表明，基于对称量测方程的多目标跟踪中，改进的遗传粒子滤波算法较扩展卡尔曼滤波算法、不敏卡尔曼滤波算法和联合概率数据关联滤波算法跟踪效果更好。