摘要
针对图像配准中存在的问题,提出云核粒子群算法。首先在云模型中采用正向正态云发生器,实现了在定性信息中获得定量的范围和分布规律;接着云滴将粒子群体分成三个子群,每个群由不同的云核进行引导,更新粒子位置与速度,云核粒子在引领子群内所有粒子到历史最佳位置的欧几里得空间距离的最大值为影响半径,对于在子群中影响半径以外的粒子,粒子核不再起引导作用;最后把图像配准过程中目标函数达到最小值问题转化为云核粒子群算法寻优问题,实验仿真得出配准后图像的满足视觉要求,位置与参考图像保持了一致。
Aiming at the existing problem in the image registration, the core particle swarm algorithm is put forward. Firstly using the forward normal cloud generator in the cloud model, quantitative scope and distribution of droplet par- ticle group are realized in the qualitative information; Then the particle swarm is divided into three subgroups, each group consists of different cloud nucleus guide, the particle position and velocity are updated. The maximum Euclide- an space distance that all the particles of the subgroup reach the historical optimum position is radius of influence. For the particles outside radius of influence in the subgroup, particle core no longer guides; Finally objective function minimum value problem in the image registration process is transformed into a cloud of nuclear particle swarm algo- rithm optimization. Simulation experiments show that the registration images can satisfy the visual requirements, posi- tion and reference images maintain consistent.
出处
《激光与红外》
CAS
CSCD
北大核心
2013年第8期951-955,共5页
Laser & Infrared
基金
河南省教育厅科学技术研究重点项目(No.13A520534)资助
关键词
云核
欧式距离
图像配准
cloud nucleus
euclidean distance
image registration
