视频内容结构化技术的研究与实现

发布时间：2018-04-10 23:34

本文选题：视频内容 + 视频结构化　；参考：《中国人民解放军国防科学技术大学》2001年博士论文

【摘要】： 视频媒体的无结构性是阻碍新一代视频应用的瓶颈问题。本文全面而系统地研究了视频无结构性问题的解决之道。在对视频制作和感知理论系统研究的基础上，将语义标注和句法分段技术有机的结合起来，在低中高三个不同的层次上对视频内容结构化技术作了深入的研究，主要的工作和贡献包括： ◆ 厘清了视频内容结构化这一交叉研究领域，明确了要解决的根本问题，应采取的技术途径和基本方法论，提出了由句法分段和语义标注构成的低、中、高整体研究框架； ◆ 系统研究了视频内容制作和感知的一般过程，指出视频内容的制作和感知实质上遵循共同的基本规律，剖析和总结了这些基本规律对视频内容结构化技术的指示意义； ◆ 全面探索了中、低层视频内容结构化技术的各种重要技术，设计和改进了部分关键算法，包括：视音频镜头、故事单元、音频大段等句法分段技术和闪光探测、音频镜头分类、段落分界符、广告、动作型故事单元等语义标注技术； ◆ 分别研究了新闻、体育、故事片和电视剧等类视频的高层内容结构化技术，解决了新闻故事探测、半透明字幕探测、角色标注、精彩标注等关键问题； ◆ 设计和实现了支撑视频内容结构化研究的VideoCar实验床，，该实验床在实际使用中收到好的效果。
[Abstract]:The lack of structure of video media is the bottleneck of the new generation of video applications.In this paper, the solution to the problem of no structure in video is studied comprehensively and systematically.Based on the systematic research of video production and perception theory, this paper combines semantic tagging with syntactic segmentation technology, and makes a deep research on the structure technology of video content at three different levels of low, middle and high levels.The main work and contributions include:This paper clarifies the cross-research field of video content structure, clarifies the fundamental problems to be solved, the technical approach and basic methodology to be adopted, and puts forward a low, medium and high overall research framework composed of syntactic segmentation and semantic annotation.This paper systematically studies the general process of video content production and perception, points out that the production and perception of video content actually follow the common basic laws, and analyzes and summarizes the instructive significance of these basic laws to the structured video content technology.This paper comprehensively explores all kinds of important techniques of middle and low level video content structure technology, designs and improves some key algorithms, including: video and audio lens, story unit, audio segment and other syntactic segmentation techniques and flash detection.Audio shot classification, paragraph delimiter, advertising, action story unit and other semantic tagging technology;This paper studies the high-level content structure technology of news, sports, feature film and TV series, and solves the key problems, such as news story detection, translucent subtitle detection, role tagging, wonderful annotation and so on.The VideoCar experimental bed which supports the research of video content structure is designed and implemented. The experimental bed has good effect in practical use.
【学位授予单位】：中国人民解放军国防科学技术大学
【学位级别】：博士
【学位授予年份】：2001
【分类号】：TN948.12

【引证文献】