摘要
针对粒子滤波跟踪算法中粒子多样性退化问题,将改进的遗传算法应用到粒子重采样中,改善了样本的多样性。在改进的遗传算法中,使用了多项式重采样进行优选复制;以特定区间的随机数做交换率进行样本交叉繁殖;使用了马尔可夫链蒙特卡罗移动加高斯白噪声做样本变异繁殖并使用快速MH抽样算法选取样本。改进后的粒子滤波跟踪算法不但保持了较高的运算效率,而且还较好地提高了跟踪的稳定性。试验表明,改进后的粒子滤波跟踪算法目标跟踪更加稳定,目标定位更加准确。
To solve the problem of particle degradation in particle filter algorithm is used in the particle re-sampling to improve the diversity of the tracking algorithm, the improved genetic sample. In the improved genetic algorithm, multinomial re-sampling is applied in selection and copy. An exchange rate of sample is a random number in given interval in cross-breeding. The Markov chain Monte Carlo move plus Gaussian white noise is used in sample variance and breeding, and the MH sampling algorithm is used to select sample, too. The improved particle filter tracking algorithm not only keeps the high efficient operation, but also improves stability of target tracking. Experimental results show that particle filter tracking algorithm is more stable and more accurate than traditional particle filter tracking algorithm.
出处
《光电工程》
CAS
CSCD
北大核心
2010年第10期16-22,共7页
Opto-Electronic Engineering
基金
国家863高技术计划资助项目(2006AAJ103)
关键词
遗传算法
粒子滤波
快速MH抽样
多模板融合
目标跟踪
genetic algorithm
particle filter
fast MH sampling
multi-template fusion
target tracking
