摘要
为自动调节当前检索任务以使最终查询结果朝着有利于用户要求的方向发展,提出一种基于组合特征双重加权的相关反馈算法。将图像检索中初始权重的设定过程作为优化问题,利用量子遗传算法求取全局最优解,作为图像检索过程中各特征初始权重的加权值;另外,在组合特征权重动态调节的过程中,将灰色关联分析理论中的灰关联度作为特征权重的估计值,同时将反馈结果中每幅图像的评价都考虑到灰色关联分析的计算中,从而来估计不同特征在检索中的相对重要性。实验结果表明,本文算法能够达到精炼检索结果的目的,大幅提高检索全面性和检索准确度。
In order to adjust the retrieval task for accommodating the query results to user＇s quest, a relevance feedback algorithm based on dual weighted combined feature is proposed in this paper. Firstly, the setting of initial weight in retrieval is conducted as optimization problem. The global optimum solution obtained from quantum genetic algorithm is used as weighted value of initial weight of each feature. Then in the process of dynamic modulating combined feature weight, grey relation degree is used as estimated value of feature weight. Also the evaluation of each image in feedback results is calculated into grey relation analysis that estimates the relative importance of different features in retrieval. The experimental results indicate that the proposed algorithm can achieve the aim of refining retrieval results, and can enhance retrieval comprehensiveness and retrieval precision largely.
出处
《信息与电子工程》
2011年第4期491-496,共6页
information and electronic engineering
基金
中央高校基本科研业务费专项资金资助项目(HEUCFR1117)
哈尔滨市科技创新人才研究专项资金资助项目(2011RFXXG028)
关键词
组合特征
相关反馈
量子遗传
灰关联度
combined feature
relevance feedback
quantum genetic
grey relation degree
