一种基于图像内容的对应关系在语义检索中的应用  

The Application of Corresponding Relationship Based on Contents of the Image in Semantic Retrieval

作　　者：曹海[1] 彭龑[1] 

机构地区：[1]四川理工学院,计算机学院,四川自贡643000

出　　处：《湘潭大学自然科学学报》2017年第2期122-126,共5页Natural Science Journal of Xiangtan University

基　　金：四川省教育厅科研项目(10ZC084);人工智能四川省重点实验室开放基金项目(2010RY006)

摘　　要：针对基于图像高层语义的检索方法问题,提出了一种基于自然场景图片语料库下的图片本体检索框架.该方法通过考虑图像内容之间的对应关系,引入人类正常图像检索过程.利用对图像内容的分块语义注释的描述,该框架解决了目前图像检索中的问题.构建好的领域本体可以用来完成定性语义描述和检索,实现了包含语义概念和空间关系的自然语言查询.最后,对该应用的有效性进行了评估,实验结果表明,该框架在检索的精确度和相关度上都表现得非常好.In order to solve the problem of image semantic retrieval based on high level image,this paper proposes an image ontology retrieval framework based on natural scene image corpus.In this method,the human normal image retrieval process is introduced by taking into account the correspondence between image contents.The proposed architecture addresses the problem of image retrieval through the use of image block representations over semantic annotation.The domain ontology can be used to complete qualitative semantic image descriptions and retrieval,and achieve natural language queries containing semantics concepts and spatial relationships.Finally,this paper evaluates the effectiveness of the architecture,and the experimental results show that,this framework performs very well in retrieval accuracy and relevance.

关 键 词：图像检索 图片本体 领域本体 语义概念 自然语言查询 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]