基于聚类的视频分类方法研究

发布时间：2019-06-17 08:43

【摘要】：伴随着互联网技术,计算机技术的快速发展,在移动通信终端和互联网平台及其他信息源产生了大量的多媒体数据。视频是一类内涵丰富、数据量庞大的多媒体数据,由此而产生了诸如视频索引、视频检索、视频分类等应用。视频分类是计算机视觉领域中的一个基础研究课题,是解决视频标注、视频检索的重要中间环节,也是进行多媒体数据管理的重要方法和手段。因此视频分类具有很高的研究和应用价值。视频是语义丰富的多媒体数据,物理特征相似的视频总有一定的相关性。本文在深入分析视频结构的基础上,将视频分割成镜头序列,根据视频所含镜头的相似性实现视频分类。本文主要做了以下研究工作:(1)深入分析了常见的视频分类器,比较了它们之间的差异,在此基础上分析并讨论了常见的视频特征提取方法。(2)提出了一种改进的K-Means聚类算法,以标签样本形成K-Means算法的初始聚类中心,以颜色特征、纹理特征和SIFT特征形成特征组来对视频进行分类,并且和一些常见的分类算法进行了比较,实验表明本文算法有较高的分类正确率及分类精度。本文的研究成果可以广泛的用于视频索引、视频检索、视频分类等应用中。
[Abstract]:With the rapid development of Internet technology and computer technology, a large number of multimedia data have been produced in mobile communication terminals, Internet platforms and other information sources. Video is a kind of multimedia data with rich connotation and huge amount of data, which results in applications such as video index, video retrieval, video classification and so on. Video classification is a basic research topic in the field of computer vision. It is an important intermediate link to solve video tagging and video retrieval, and it is also an important method and means for multimedia data management. Therefore, video classification has high research and application value. Video is a kind of multimedia data with rich semantics, and video with similar physical characteristics always has certain correlation. In this paper, based on the in-depth analysis of the video structure, the video is divided into shot sequences, and the video classification is realized according to the similarity of the shots contained in the video. The main work of this paper is as follows: (1) the common video classifiers are analyzed deeply, the differences among them are compared, and the common video feature extraction methods are analyzed and discussed. (2) an improved K-Means clustering algorithm is proposed to form the initial clustering center of K-Means algorithm with label samples, and the video is classified by color feature, texture feature and SIFT feature group. Compared with some common classification algorithms, experiments show that the proposed algorithm has higher classification accuracy and accuracy. The research results of this paper can be widely used in video index, video retrieval, video classification and other applications.
【学位授予单位】：淮北师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】