融合类无标度粒子群算法的图像检索研究  

Research on Image Retrieval Based on Fusion the Scale Free Particle Swarm Optimization

作　　者：靳雁霞[1] 薛丹 张鑫[1] 李伟楠 

机构地区：[1]中北大学计算机与控制工程学院,山西太原030051 [2]太原市气象局,山西太原030051

出　　处：《微电子学与计算机》2018年第1期36-40,共5页Microelectronics & Computer

基　　金：国家自然基金(61379080)

摘　　要：提出一种融合类无标度网的禁忌粒子群(ST-PSO)算法,并用其优化改进后的图像检索相关反馈过程.该方法通过融合类无标度网的禁忌粒子群算法优化反馈过程,并根据用户的反馈信息引入分辨率来动态调整特征权重以及查询向量.通过实验验证了该方法能够增加粒子种群多样性,避免早熟收敛,增强粒子跳出局部最优解的能力,并且可以检索到更加符合用户需求的图片,有效地提高了检索效率和检索精度.This paper propose a Tabu search particle swarm algorithm which fusions the Scale Free Like network to optimize the improved image retrieval relevance feedback. This method opitimizes the feedback process by the Tabu search particle swarm algorithm which fusions the Scale Free Like network and lead into resolution ratio through user＇s feedback information to adjust the feature weight and the query vector dynamically. Experiment shows that the method can not only increase the diversity of particle population, avoid premature convergence, enhance the ability of particle to jump out of local optimal solution, but also retrieve pictures that more user-friendly, effectively improved the retrieval efficiency and retrieval accuracy.

关 键 词：粒子群算法 无标度网 图像检索 特征权重 相关反馈 

分 类 号：TP180[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]