摘要
针对视频高层语义分析问题,文章结合足球比赛的领域知识,按照足球比赛转播、视频编辑的一般规律,根据足球比赛语义事件随机性的特点,选择特定的视频物理特征,应用HMM(隐马尔科夫模型)分析视频的语义结构,确定视频和HMM模型中各元素的对应关系,构建一个基于HMM的视频语义分析框架,并通过进行足球视频HMM参数的训练,得到视频各语义事件的HMM模型,达到视频语义自动分析的目的。
Aimed at the problem of video semantic analysis, semantic analysis framework of soccer video is presented by combining domain knowledge ofsoccer game, considering game broadcasting and editing ofsoccer video, and random characters ofgame semantic events, then given features of soccer video is choosen, and video semantic structure is analyzed by applying HMM, therefore corre- sponding relation of correlative factors between soccer video and hidden Markov model is established. Under the framework, the HMM of each semantic event is established by training HMM parameters of the soccer video, to achieve the aim of analyzing video semantic events automatically.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第19期5002-5005,共4页
Computer Engineering and Design
基金
广州体育学院自然科学基金项目(06QN06)