基于相关反馈地文物图像语义标注研究.pdfVIP

基于相关反馈的文物图像语义标注研究 耿国华韦娜黄世国 (西北大学信息科学与技术学院陕西西安719069) 摘要：“语义鸿沟”问题是目前图像检索领域内的研究热点。本文分析了图像 语叉标注的现状夏难点。提出了基于相关反馈的文物语义标注方法。在标注过程中 介入人机交互，以此采改进标注的准确率与效率。算法首先通过相关反馈构建语义 相似性矩阵，进而通过语义聚类实现语义标注。在对文物图像的标注试验中，获得 了较好的标注准确率与效率。 关键词：文物图像；语义标注；相关反馈 on ResearchSemanticAnnotationtoCulturalRelic Images BasedonRelevanceFeedback GuohuaWeiNa Geng HuangShiguo ofInformationScienceand (College Technology，NorthwestUniversity，Xi’nil Shaanxi，710069) Abstract￡“Semanticof interestto inthefield gap”isgreat theresearchers of retrieval this theresearch andthe situation in image currently．Inpaper difficulty semanticannotationareoverviewed．Anannotationmethodbasedon image image therelevancefeedbackiS iS forward．Man-machineinteractionintroducedinthe put is constructed annotation．Thebasedonthe matrix image algorithm similarity by therelevance thesemantic is to theseman— feedback，then clusteringperformedget tic resultsona collectionof haveshownthe annotation．Experimental large images and effectivenessrobustnessofour proposedalgorithm．