视频语义特征提取技术
视频语义特征提取技术(任务书,开题报告,外文翻译,论文18000字)
摘要
目前互联网技术飞速发展，多媒体数据也得到了极大的传播，我们每天都在面对众多的信息。然而目前的检索技术不能让我们很好的利用起这些信息，很多的检索方式都是依据文本标注的方式进行的，基于内容的视频检索系统也更多是利用低层特征。如何在高层语义上检索视频，跨越语义鸿沟成为研究的难点。
本文分析了目前视频检索技术的现状，基于内容的视频检索现有的成果和存在的问题，研究了镜头分割和关键帧的提取，对于关键帧低层特征的颜色和纹理提取，主要利用颜色直方图，颜色矩和灰度直方图。在低层特征到高层语义映射中分析了KNN算法，朴素贝叶斯算法和高斯核算法，并分析了其优缺点。在文章最后提出了缩小语义鸿沟，可在镜头的分割，增加多样低层特征以及考虑语义频率对结果的影响等问题上研究。
关键词：QBIC；高层语义；低层特征；关键帧；语义标注
Abstract
At present, the development of Internet technology is quick.What is more,multimedia data has also been spread in a large extent, which means that we are facingvarious kinds of information every day. However, the current video retrieve technology could not let us make good use of this information.A lot ofvideo retrieve technology is based on textannotation,content-based video retrieval systems also using low-level features to retrieve videos. How to retrieve video in high-level semantics and cross the semantic gap becomes the most difficult part of research. [资料来源：https://www.doc163.com]
This paper analyzes the existing results and existing problems of content-based video retrieval. The extraction of the camera lens and the extraction of the key frame are studied. For the color and texture extraction of the low-level features of the key frame, the color histogram, the color moment and the gray histogram are mainly used. The KNN algorithm, the naive Bayesian algorithm and the Gaussian kernel algorithm are analyzed in the low-level feature to the high-level semantic mapping, and the advantages and disadvantages are analyzed. At the end of this paper, I propose some ways of narrowing the semantic gap, which can be done by studying on the division of the camera lens, adding various low-level features and considering the influence of semantic frequency on the results.
Keywords:QBIC; high-level semantics; low-level features; key frames; semantic annotation
[资料来源：https://www.doc163.com]
目录
第1章绪论 1
1.1引言 1
1.2视频检索的背景和发展 1
1.2.1基于内容的视频检索提出背景 2
1.2.2现有的系统 2
1.2.3存在的问题 3
1.2.4基于语义的视频检索研究现状 4
1.3研究的目的和意义 4
1.4课题研究内容 5
第2章视频镜头分割和关键帧的提取 6
2.1镜头分割 6
2.1.1基于像素比较方法 7
2.1.2直方图方法 7
2.1.3镜头分割的难点 7
2.1.4突变渐变自检测改进算法 8
2.2关键帧的提取 10
2.2.1基于镜头边界的方法 10
2.2.2基于视觉内容的方法 10
2.2.3基于运动的方法 11
第3章低层特征的提取 12
3.1颜色特征 12
3.1.1颜色空间 12
3.1.2颜色直方图 14
3.1.3颜色矩 14
3.2纹理特征 15
3.3形状特征 16
第4章低层特征到高层语义建立映射的算法 17
4.1K近邻算法 17
4.2朴素贝叶斯算法 18
4.2.1.分类问题 18
4.2.2.贝叶斯定理 19
4.2.3利用朴素贝叶斯分类进行语义标注 19
4.3高斯核算法 20
4.4对各种算法的比较和分析 21
第5章实验和分析 23
第6章总结与展望 30
参考文献 32
致谢 33 [版权所有：http://DOC163.com]