羽毛球视频自动剪辑技术研究

Research on Badminton Video Automatic Editing Technology

当前针对羽毛球比赛视频的自动剪辑技术尚不完善,该文通过分析完整的羽毛球比赛视频,将其分为比赛片段、回放片段与人物特写片段,并采用图像处理技术实现三类片段的完整剪辑。该文针对比赛片段提出了结构相似度判断法与霍夫直线检验法,针对回放片段采用了SIFT特征匹配法与轮廓检测法,针对人物特写片段则采用了深度学习方法,并对各种方法的实验结果进行评价和分析,为羽毛球视频自动剪辑系统的实现提供了理论基础。

The current automatic clip technology for badminton match video is not perfect,this paper analyzes the complete badminton match video and divides it into match segments,playback segments and close-up segments,and uses image processing technology to achieve the complete clip of three types of clips.In this paper,we propose the structure similarity judgment method and hough line test in view of the game pieces,we adopt the SIFT feature matching method and contour detection for playback footage,we use the deep learning method for characterization fragment,and we also evaluate and analysis the experimental results of various methods,providing the implementation of badminton automatic video editing system with theoretical basis.