计算机应用 ›› 2005, Vol. 25 ›› Issue (07): 1651-1653.DOI: 10.3724/SP.J.1087.2005.01651

• 人工智能 • 上一篇    下一篇

一种基于PBIL算法的快速图像匹配方法

宋晓宇,刘云鹏,王永会   

  1. 沈阳建筑大学  信息与控制工程学院
  • 收稿日期:2004-12-17 修回日期:2005-03-08 发布日期:2005-07-01 出版日期:2005-07-01
  • 作者简介:宋晓宇(1963-),男,山东招远人,教授,博士,主要研究方向:图像处理、GIS;刘云鹏(1980-),男,河南汝南人,硕士研究生,主要研究方向:图像处理;王永会(1970-),男,黑龙江尚志人,副教授,主要研究方向:数据库、图像处理
  • 基金资助:

    〗国家科技成果重点推广项目(2004EC000096)

A  fast image matching method based on PBIL algorithm

SONG Xiao-yu,LIU Yun-peng,WANG Yong-hui   

  1. Faculty of Information & Control Engineering, Shenyang Jianzhu University
  • Received:2004-12-17 Revised:2005-03-08 Online:2005-07-01 Published:2005-07-01

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。

关键词: PBIL算法, 图像匹配, 相关匹配, 遗传算法

Abstract:

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.

Key words: PBIL algorithm, image matching, correlation matching, genetic algorithm

中图分类号: