摘要
针对TBVR技术中人工标注存在的问题和CBVR技术的不成熟,以及对虚拟新闻系统中视频检索需求和特点的深入分析,在TBVR的基础上提出了一种基于多关键字的新闻视频自动检索方法。详细讨论了标注字典库、树形标注结构、关键字自动获取、相似度计算模型和自动检索算法等问题,并进行了实验验证。结果表明,提出的方法在对新闻视频进行检索时取得了很高的查全率,同时取得了较高的查准率,能够解决虚拟新闻系统中视频自动检索的问题。
Aiming at the problem of manual annotation in TBVR and the immatureness of CBVR and in-depth analysis on the requirements and features of video retrieval in virtual television news, a news video retrieval method based on multiple keywords was proposed. Annotation dictionary, tree annotation structure,automatic keywords extraction, similarity computation model and automatic retrieval algorithm were discussed in detail. An experiment was made on this method. The result shows that it has attained a very high recall rate with a comparative high accuracy rate and it can solve the problem of automatic video retrieval in virtual television news.
出处
《计算机科学》
CSCD
北大核心
2010年第1期181-183,188,共4页
Computer Science
基金
国家863项目(编号:2006AA01Z337)
国家973项目(编号:2006CB303106)资助
关键词
虚拟新闻
视频检索
视频相似度
战略对抗演习
Virtual television news, Video retrieval, Video similarity, Strategic seminar gaming