一种基于PBIL算法的快速图像匹配方法  

A fast image matching method based on PBIL algorithm

作　　者：宋晓宇[1] 刘云鹏[1] 王永会[1] 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168

出　　处：《计算机应用》2005年第7期1651-1653,共3页journal of Computer Applications

基　　金：国家科技成果重点推广项目(2004EC000096)

摘　　要：为了解决图像匹配过程中计算速度慢和匹配精度不高的缺陷,提出了一种基于群体增量学习算法的匹配方法。PBIL算法是一种基于概率分析的进化算法。它集成了基于函数优化的遗传搜索和竞争学习两种策略,将进化过程视为学习过程,通过竞争学习所获得知识来修正生成概率,进而指导后代的生成。在实验中,将其与传统序贯相似性检测算法(SSDA)和遗传算法进行了比较。结果表明基于该算法的图像匹配具有运算速度快、匹配精确等优点,且收敛过程非常稳定。To solve the problem of slow computation speed and low image matching accuracy, a new approach to image matching using population-based increased learning algorithm (PBIL) was proposed. PBIL algorithm is a probability learning based evolutionary algorithm. It integrates genetic search strategy based on function optimization with competitive learning strategy. It regards evolution as a learning process, and revises the produce probability of offspring according to knowledge come from competitive learning. Compared with the conventional sequential similarity detection algorithm and genetic algorithm, the experiment results show that this approach is fast in operation, and has high accuracy in matching, and the convergence is very stable.

关 键 词：PBIL算法 图像匹配 相关匹配 遗传算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]