权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

A Study on Automatic Indexing of Image Contents Based on Semantic Field by Modal Logic

基于语义场的模态逻辑图像内容自动索引研究

基本信息

批准号：
14380171
负责人：
MURAI Tetsuya
金额：
$ 9.47万
依托单位：
HOKKAIDO UNIVERSITY
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
2002
资助国家：
日本
起止时间：
2002 至 2005
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-14380171/
关键词：
modal logic semantic field image databases Kripke models Scott-Montague models automatic indexing content retrieval filtration 中間媒体自動牽引付け

项目摘要

The final purpose of this research project is to provide an automatic indexing system and its application to image content retrieval. When we consider image retrieval and indexing, we must consider two levels of both logical (linguistic) and physical (image). To connect the two levels we need some intermediate media, as which we adopt the theory of semantic field by modal logic, which was formerly proposed by Professor Germano Resconi from Italy and myself. The semantic field gives us a way of visual treatment of logical inference. The project contains the following two steps :(1)Indexing by semantic field as intermediate media and its application to image content retrieval(2)Automatic indexing of image content by logical language.For the purpose we develop a new theory of granularization based on semantic field in the project. The concept of granurality has recently put emphasis on in the research field of rough sets, which are algebraic counter parts of modal logics. In our theory of granularization, we define two mathematical operations of zooming in and out in a modal logical setting and then apply them to adjustment of granularity, which has an important role in image processing. The idea is originated in the so-called filtration method in modal logic, which is the basic way of decision problems. Then by regarding images as semantic fields on 2-dimensional Euclidean space, we can apply the two operations to images. As a result we can develop an image content retrieval system and make the basis of automatic indexing of image contents by a logical language. It is very important that we can point out that our way of indexing is very close to abductive reasoning.

本研究的最终目的是提供一个自动索引系统及其在图像内容检索中的应用。当我们考虑图像检索和索引时，我们必须考虑逻辑（语言）和物理（图像）两个层面。为了连接这两个层次，我们需要一些中间媒介，作为这一媒介，我们采用了模态逻辑的语义场理论，这一理论是由意大利的Germano Resconi教授和我本人提出的。语义场为我们提供了一种直观地处理逻辑推理的方法。该项目包括以下两个步骤：（1）语义场作为中间媒介的索引及其在图像内容检索中的应用（2）基于逻辑语言的图像内容自动索引。粒度的概念是近年来在粗糙集研究领域中的一个热点，粗糙集是模态逻辑的代数对应物。在我们的粒度化理论中，我们定义了两个数学运算的放大和缩小在一个模态逻辑设置，然后将它们应用到调整的粒度，这在图像处理中有重要的作用。这一思想起源于模态逻辑中所谓的过滤方法，它是解决决策问题的基本方法。然后将图像看作二维欧氏空间上的语义场，将这两种操作应用于图像。从而可以开发一个图像内容检索系统，为用逻辑语言实现图像内容的自动标引奠定基础。非常重要的是，我们可以指出，我们的索引方式非常接近溯因推理。