图像信息资源可视化协同语义标注及实现研究

项目名称： 图像信息资源可视化协同语义标注及实现研究

项目编号： No.71273195

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 管理科学

项目作者： 陆泉

作者单位： 武汉大学

项目金额： 54万元

中文摘要： 随着图像信息资源的迅速增长，图像标注中存在的语义鸿沟问题严重影响了图像信息资源研究与应用的发展，将人对图像的感性认知与自动图像标注有机联系起来协同标注是解决问题的有效途径。本项目综合应用情报学、管理学、计算机科学、认知科学等多学科的理论与方法，首先通过理论分析与用户认知实验，深度剖析图像语义鸿沟的产生过程，明确图像标注的研究需求；进而构建图像信息资源的可视化协同语义标注模型，以图像与文本协同挖掘进行图像标签语义优化，以语义可视化支持用户感性交互，通过用户对图像及图像之间关系的感性认知与计算机语义处理之间的协同，研究消减乃至消除语义鸿沟的图像标注；接着以典型数据集及web实际数据从系统与用户体验两方面对模型进行实证研究；最后探索将模型应用于图像信息资源研究与实践的途径与方式，为有效标注大规模图像信息资源的多层次语义提供理论与应用指导，并将丰富与拓展图像信息资源管理理论与方法。

中文关键词： 图像标注；语义鸿沟；可视化；协同；

英文摘要： With the rapid growth of image information resources, semantic gaps in existing image annotation theories including Automatic Image Annotation(AIA) have hindered the development in research and application of image information resource field seriously, and an effective way to narrow them is to link up the perceptual-cognition of user to image with AIA organically to annotate images collaboratively. This project integrates application of theory and methods from multi-disciplinary field, including Information Science, Management, Computer Science, and Cognition. First, by theoretical analysis and user cognitive experiments, the production process of image semantic gap is deeply analyzed, which states clearly the needs in image annotation research. Then a cooperative annotation model of images via semantic visualization is presented to solve the problem of semantic gap，in which collaborative mining methods based on images and their text information are studied to assist users by semantic optimization and two semantic visualization models of image information resources are built to enable user to annotate images only with his/her perceptual-cognition to them. This model aims to annotate images avoiding semantic gap through the collaboration of user and computer. Typical data sets and a web data set about image annot

英文关键词： image annotation；semantic gap；visualization；cooperation；