图像检索中基于粒子群优化的相关反馈算法  

Particle Swarm Optimization Based Relevant Feedback Algorithm in Image Retrieval

作　　者：许相莉[1] 张利彪[1] 刘向东[1] 于哲舟[1] 周春光[1] 

机构地区：[1]吉林大学计算机科学与技术学院,吉林长春130012

出　　处：《小型微型计算机系统》2009年第5期971-973,共3页Journal of Chinese Computer Systems

基　　金：国家自然科学基金重点项目(60433020;60673099;60773095)资助;国家"八六三"计划项目(2007AA04Z114)资助;"九八五"工程:"计算与软件科学科;技创新平台"项目支持;欧盟项目(TH/AsiaLink/010(111084))资助;教育部重点(02090)资助;吉林省科技发展计划项目(20080168)资助;吉林大学"九八五"工程研究生创新基金项目(20080236)资助

摘　　要：查询优化问题虽然整体的优化目标是明确的,但由于用户主观理解导致优化目标的细节会有一定的随机性,并且由于特征提取算法中语义鸿沟的存在,导致优化方向的不稳定,本文利用粒子群优化(particle swarm optimization,PSO)算法全局寻优、快速收敛的优点,针对查询优化中相关反馈(relevant feedback,RF)问题的特点,提出一种理想值监督下的粒子群优化-相关反馈(PSO-RF)算法,算法在理想值监督下指导粒子的运动方向,进而快速向理想解集靠拢.实验表明,该方法能够有效的提高检索性能,取得较好的检索结果.Although the global optimization objective of query optimization problem is clear, the users- subjective understanding induces certain randomicity for the details of optimization objective,and because the existing of semantic gap in feature extraction algorithms, the optimization direction is instability. In this paper, the excellences of particle swarm optimization（PSO） algorithm that global optimization and fast convergence were utilized, and aiming at the characteristics of relevant feedback（ RF）, an PSO-RF algorithm under ideal value supervised is proposed, which could guide the movement direction for particles, and close to ideal solution set quickly. Experiments proved that the algorithm could improve retrieval capability well, and has achieved better retrieval result.

关 键 词：粒子群优化 理想值监督 相关反馈 图像检索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]