【摘要】 针对自动视频分类工作中分类预测精度低的问题,提出了一种集成数据挖掘技术的自动视频分类方法。首先进行视频分割,形成了一个视频属性数据库;然后分别使用决策树、分类关联规则等技术对视频属性数据库进行数据挖掘,提取出决策树分类规则集和分类关联规则集;最后利用一个规则集的合并裁减算法来合并这两个分类预测规则集,形成最终的具有更高精度的视频分类规则集。通过实验验证了决策树分类预测规则和分类关联规则具有分类预测的一致性;同时实验表明,使用合并后的规则集比单独使用一个规则集来预测视频具有更高的预测准确率。更多还原

【Abstract】 This paper addresses the problem that automatic videoclassification has lower classification precision. An automatic video classification approach of integrating data mining is proposed. The approach firstly uses video segmentation to extract a set of video features and construct a video attribute database. Then the decision tree and class association rules mining are used on the video attribute database to extract a decision tree classification rule set and a class association rule set respectively. Lastly this two sets of rule extracted are combined by an algorithm for combining two rule sets to generate the ultimate classification rule set that has higher classification accuracy. The experiment result verifies the consistency of decision tree classification with class association classification. The result also shows that the combined rule set has higher classification precision than just using one rule set.更多还原