CAJ | 학술논문

重采样是解决粒子滤波退化问题的主要方法，重采样的基本思想是采取复制保留权值较高的粒子，删除权值较低的粒子，而这导致了粒子多样性的减弱，特别是在样本受限条件下，甚至导致滤波发散。针对上述问题，提出改进的粒子滤波算法，将Mean Shift与粒子滤波融合，在重采样部分引入小生境遗传算法，提高粒子的多样性，避免粒子退化。实验表明，改进后的算法状态估计精度更高，效果更好。
중채양시해결입자려파퇴화문제적주요방법，중채양적기본사상시채취복제보류권치교고적입자，산제권치교저적입자，이저도치료입자다양성적감약，특별시재양본수한조건하，심지도치려파발산。침대상술문제，제출개진적입자려파산법，장Mean Shift여입자려파융합，재중채양부분인입소생경유전산법，제고입자적다양성，피면입자퇴화。실험표명，개진후적산법상태고계정도경고，효과경호。
Resampling is a critical operation to solve degeneracy problem with particle filters generally. The basic idea of resampling is to discard particles which have small weights and concentrate on particles with large weights. But resampling often introduces sample impoverishment problem, especially the sample is limited under the condition, even causes the filter to disperse. This paper proposes improved particle filter algorithm. Mean Shift integrates with particle filter, and then the niching genetic algorithm is used in resampling in order to improve the variety of particles and remove the degeneracy phenomenon. The simulation results prove the proposed algorithm reduces the tracking error, and has better precision.