摘要
针对传统的粒子滤波跟踪算法存在粒子退化的问题,提出了一种结合人工免疫的粒子滤波跟踪算法。该方法利用免疫学原理,将目标模板特征作为抗原,每个粒子对应区域的特征作为抗体,匹配问题转化为抗原和抗体的亲和力问题,通过克隆的方式保留亲和力大的抗体,采用变异的手段去除亲和力小的抗体,从而使结果快速收敛于全局最优解。抗体的多样性有效解决了传统粒子滤波的退化问题。将该方法应用到目标跟踪技术中,仿真结果表明,粒子集的有效样本得到了明显的提高。
For the traditional particle filter tracking algorithm with the problem of particle degeneration,a particle filter tracking algorithm combining artificial immune is presented.The method utilizes the principle of immunology, makes the feature of target template as antigen and the feature of every particle＇s corresponding area as antibody, transforms the problem of matching into the affinity between antigen and antibody, keeps the antibody with high affinity by the way of clone and gets rid of the antibody by means of variation,thus the result converges at the global optimal solution rapidly.The diversity of antibody solves the degeneration of traditional particle filter effectively.Using the method in target tracking,the simulation＇s resuits show that the effective sample of particle set is improved apparently.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第28期195-197,222,共4页
Computer Engineering and Applications
关键词
粒子滤波
人工免疫
粒子退化
目标跟踪
有效样本
panicle filter
artificial immune algorithm
particle degeneration
target tracking
effective sample
