面向技战术分析的足球视频中的运动模式挖掘

项目名称： 面向技战术分析的足球视频中的运动模式挖掘

项目编号： No.61202330

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 杨颖

作者单位： 中国农业大学

项目金额： 23万元

中文摘要： 面向足球视频技战术分析的运动信息挖掘综合利用视频处理、人工智能、模式识别及计算机视觉技术挖掘体育视频高级语义信息，具有十分重要的理论价值和应用前景。当前足球视频技战术信息挖掘存在关键运动对象选择和战术分析需求差异、运动模式挖掘的准确率和效率低及运动模式缺乏合理有效的分类识别等问题。针对上述问题，本项目开展了面向广播足球视频技战术分析的高效的运动模式提取和挖掘研究，拟提出一种基于焦点运动员分析的战术运动模式挖掘方法，采用动态规划的算法提取焦点运动员的运动轨迹，相对于现有的跟踪算法,提高了运动员检测的效率和准确率,并在此基础上构建面向真实场地的多元时序进攻轨迹作为高效的战术内容表达，最后提出一种多尺度曲线拟合的运动模式分类和识别方法，实现对不同层次战术内容的识别分析,为体育专业人士指导比赛训练、提高竞技水平提供科学指导，满足互联网用户的多层次视频检索和观赏需求，促进多媒体内容检索技术的发展。

中文关键词： 足球视频；运动模式；焦点运动员；进攻轨迹；战术分析

英文摘要： Motion pattern mining for soccer tactics analysis extracts the higher semantic content in soccer videos by integrating the techniques of video analysis, artificial intelligence, pattern recognition and computer vision, and has very important research values and wide application perspectives. The current work of tactics content extraction for soccer video has many problems, including the gap between the moving object selection and tactics analysis requirement, the low accuracy and efficiency of motion pattern mining and the lack of reasonable recognition and classification of motion patterns. To address the problems mentioned above, this project studies the efficient and accurate mid-level semantic content extraction and motion pattern mining for tactics analysis in broadcast soccer videos. We plan to propose a novel method of motion pattern mining based on focus player detection for tactics analysis, which extracts the positions and trajectory of focus player based on dynamic programming and achieves better performance compared with the tracking algorithms. Based on the extracted trajectories of focus players, a multivariate time series attack trajectory is constructed for efficient representing the tactic contents. Finally, we intend to propose a multi-scale curve fitting based method for motion pattern classif

英文关键词： sports video；motion pattern；focus player；attack trajectory；tactics analysis