基于Q学习的新闻图像检索方法

摘要

针对新闻图像检索的应用特点,提出了一种多反馈、合作型的图像检索方法.通过构造动态的Q表,保存图像的折算累计反馈；设计从探索型逐渐过渡到利用型的图像选择策略；在方差分析的基础上,设计了多反馈综合方法,全面地获取用户检索需求,从而构造了基于Q学习的相关反馈检索算法.实验结果表明了该算法是有效的,并具有更高的性能.%According to application features of news image retrieval, a multi-feedback and cooperative retrieval method is proposed. A dynamic Q-chart is constructed to store discounted cumulative rewards of images. Image selecting strategy is designed transited from exploring to exploiting. And, a multi-feedbacks synthesizing method is constructed at the basis of variance analysis to gain total requirement of user. So, a relevance feedback algorithm based on Q-learning is proposed. Experimental results show the proposed algorithm is effective and has higher performance.