基于颜色与边缘分布的足球视频镜头分类方法  

Soccer Video Shot Classification Method Based on Color and Edge Distribution

作　　者：周艺华[1] 曹元大[1] 李剑[2] 张洪欣[3] 

机构地区：[1]北京理工大学信息科学技术学院计算机科学工程系,北京100081 [2]北京邮电大学信息安全中心,北京100876 [3]北京邮电大学通信网络综合技术研究所,北京100876

出　　处：《北京理工大学学报》2005年第12期1079-1082,共4页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金资助项目(60271018)

摘　　要：针对足球视频镜头的特点,提出了一种基于颜色与边缘分布的视频镜头分类方法.首先利用高斯混合模型(GMM)对草地颜色建模,然后利用已建立的GMM计算出每个镜头的草地颜色分布特征.针对颜色特征对光照的敏感性,提出利用边缘分布作为颜色特征的补充,二者的结合突出了不同镜头的特点.利用所提出的特征,采用适合小样本分类问题的支持向量机(SVM)对足球视频镜头分类.实验表明,新方法的平均分类准确率可达92%以上.Classified by the characters of shots for soccer video, a color and edge distribution based shot classification method is presented. A Gaussian mixture model （GMM） is first trained for grass pixel, the grass distribution is then computed based on the trained GMM. Due to the sensitivity to lights for color features, edge distribution is computed to complement the drawbacks of color features. Integrating the two types of features embodies the main characters of different shot types. Support vector machines （SVMs） which suit to classification problem for tiny samples is designed for different shot types through the features extracted by the method. Experiments showed that the average precision of the method can attain over 92%.

关 键 词：视频检索 颜色分布 边缘分布 镜头分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]