项目名称: 基于跨媒体数据挖掘的社会图像事件分析与标注
项目编号: No.61202239
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 张小明
作者单位: 北京航空航天大学
项目金额: 25万元
中文摘要: 社会网络媒体的迅速发展对多媒体数据管理和应用提出了严峻的挑战,即如何有效地管理大量的、动态的、用户产生的多媒体数据的问题。传统社会图像事件分析和标注技术主要关注图像本身的信息,没有综合考虑不同特征的关联关系对社会图像事件检测的作用,没有利用跨媒体数据挖掘技术更深入全面的分析社会图像事件,并且没有充分利用图像的上下文信息提高图像标注算法的性能。为了充分利用社会网络媒体丰富的内容以提高社会图像管理与应用的性能,本项目拟对社会图像事件分析与标注进行研究。主要研究内容包括:融合多种特征的社会图像事件检测与跟踪;基于跨媒体数据挖掘的社会图像事件分析;结合事件信息的社会图像标注。本项目的研究将充分挖掘跨媒体事件的关联模式,并探索利用社会网络丰富的信息改进图像标注算法性能,为更全面分析社会网络事件和优化社会图像的管理提供理论依据和技术支撑。
中文关键词: 图像标注;事件检测;图像话题模型;多模数据;图像检索
英文摘要: The rapid development of social network media poses a severe challenge to the multimedia data management and application, which bring up a serious problem of how to effectively manage a large number of dynamic and user-generated multimedia data. The traditional technologies of social image event analysis and annotation mainly focus on the image itself. They do not consider the relationship between multi-type features in social image event detection, and they do not use the cross-media data mining technology to make a more in-depth and comprehensive analysis of social image events. Moreover, the context information of social image isn't well used to improve the performance of social image annotation. To take full advantage of the rich content of social networking media to enhance the social image management and application performance, this project aims to do research on event analysis and annotation of social image. The research topics include: event detection and tracking of social image based on multi-type feature fusion; event analysis of social image based on cross-media data mining; incorporating event information for social image annotation. The research of this project will fully mine the associated pattern of cross-media events, and also exploit to improve the performance of image annotation algorithms b
英文关键词: image tagging;event detection;image topic model;multi-modal data;image retrieval