演化视角下图像的语义表示
发布时间:2018-06-03 00:29
本文选题:图像语义表示 + 语义演化 ; 参考:《图书情报知识》2017年06期
【摘要】:作为记录信息的一种重要形式,图像包含了丰富的语义信息。随着图像资源的不断累积,有效的图像语义表示成为了提高图像检索效率的关键。文章梳理了关于图像描述与表示的相关研究,归纳出现有研究仍存在对过程性知识的表示相对不足,反方向的概括关联表示相对薄弱这两项问题。因此为探究自然有效且可计算的图像语义表示方式,我们提出以"演化"的视角来剖析图像的语义表示问题。文章构建了基于表示过程的层次描述框架,深入分析了该表示方式的重点与难点,并以具体案例示例说明。基于演化视角的图像语义表示重视表示结果与过程,符合用户解读习惯,为更高层次的语义推理和知识发现累积更丰富且可计算的信息。因此该图像语义表示方法具有一定的实用性,可被应用于辅助专家/用户的图像标注等具体实践领域。
[Abstract]:As an important form of recording information, images contain abundant semantic information. With the continuous accumulation of image resources, effective image semantic representation has become the key to improve the efficiency of image retrieval. This paper combs the related research on image description and representation, and concludes that there are still two problems, that is, the representation of process knowledge is relatively insufficient, and the general association representation in the opposite direction is relatively weak. Therefore, in order to explore the natural and computable image semantic representation, we propose to analyze the image semantic representation from the perspective of "evolution". In this paper, a hierarchical description framework based on representation process is constructed, and the emphases and difficulties of this representation are analyzed in depth, and an example is given to illustrate it. Image semantic representation based on evolutionary perspective attaches importance to the representation of results and processes, conforms to the custom of user interpretation, and accumulates more abundant and computable information for higher level semantic reasoning and knowledge discovery. Therefore, the method of image semantic representation is practical and can be used in practical fields such as auxiliary expert / user image annotation.
【作者单位】: 武汉大学信息管理学院;武汉大学信息资源研究中心;
【基金】:国家自然科学基金重大研究计划“大数据驱动的管理与决策研究”重点支持项目“基于知识关联的金融大数据价值分析、发现及协同创造机制”(91646206)和“大数据环境下的知识组织与服务创新”(71420107026) 教育部人文社会科学重点研究基地重大项目“大数据资源的语义表示与组织研究—面向文化遗产领域”(16JJD870002)的研究成果之一
【分类号】:G254
,
本文编号:1970764
本文链接:https://www.wllwen.com/tushudanganlunwen/1970764.html