【摘要】 为弥补特征提取中的语义缺陷,提出了一种利用领域知识规则填补特征与高级语义之间鸿沟的思想,从体育视频中对语义对象进行有效的特征提取,并采用支持向量机元分类器和组合策略对体育视频进行分类的方法.实验表明,该分类方法对大部分体育视频都具有很好的分类效果,平均准确率可达92.23%,优于其他提取特征无语义关联的分类方法.更多还原

【Abstract】 A novel SVM meta classifier based sport video classification algorithm is presented to bridge the low level feature and high level semantic feature.Domain knowledge rules are exploited to extract features semantically.Meta classifiers classify the video clips with combination strategies.The experimental results showed that the algorithm can be used in almost all sports video classification,and have better performance than other non-semantic associate classification algorithms with an accuracy attaining 92.23%.更多还原