基于多特征视频关键片段提取研究--以健康类动画视频为例  

A Multi-features-based Key Segment Extraction Model for Healthy Animation Video

作　　者：肖栩豪 李晓军 姚俊萍 李少臣 XIAO Xv-hao;LI Xiao-jun;YAO Jun-ping;LI Shao-chen(Rocket Force University of Engineering,Xi’an 710025,China)

机构地区：[1]火箭军工程大学,陕西西安710025

出　　处：《中国电子科学研究院学报》2021年第6期561-568,591,共9页Journal of China Academy of Electronics and Information Technology

摘　　要：短视频爆发式增长给内容检索和分析造成极大困难,为了提高短视频内容治理的有效性,有必要开展视频关键片段提取研究。文中提出了视频关键片段提取两阶段模型,在视频片段划分阶段,采用颜色和像素灰度双特征融合方法对视频进行片段划分;在关键片段提取阶段,首先基于Canny边缘检测的"横向压缩"定位方法对视频字幕定位,进而基于关键词词量统计确定单一关键片段。实验分析表明,文中提出的方法在片段划分基础上,能够准确提取代表视频内容的单一片段,有效压缩了视频内容数据量,为视频快速检索及内容分析提供了技术手段。The explosive growth of the number of short video has caused great difficulties in its content retrieval and analysis,in order to improve the effectiveness of short video content governance,research on the extraction of key segments in short video needs to be advanced.In this paper,a two-stage for key segment extraction of short video model is proposed.In the stage of video content segmentation,the video is segmented by a new method based on color and pixel gray level dual feature fusion.In the stage of key segment extraction,"Transverse compression"localization method based on Canny edge detection locates the subtitle of short video,and then extract a single key segment based on the key words statistics.Experimental results show that the proposed method can accurately extract a single segment representing the short video content,and effectively compress the video content data.It provides a technical means for quick retrieval and content analysis of short video.

关 键 词：短视频 关键片段 字幕检测 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]