摘要： 目前视频数据挖掘概念、体系尚不明确，现有的一些分类方法不能突出数据挖掘关于“新颖性”这一特性。该文尝试从概念和分类方法2个方面解决这些问题，根据数据挖掘和视频各自的特点，将视频数据挖掘的概念分为广义和狭义2种。从数据分析的角度，提出一种新的分类方法，将广义的视频挖掘分为描述性和预测性2类，并且按照视频内容所关联的领域，提出了一个描述性视频数据挖掘的分类框架，讨论了视频挖掘的现有方法，并展望了视频数据挖掘的发展趋势。

关键词: 视频数据挖掘, 描述性视频挖掘, 预测性视频挖掘

Abstract: Video data mining is an emerging research area and some of the existing classifications are unable to highlight the “novelty” nature of the data mining. This paper attempts to clarify these problems. According to the characteristics of data mining and video sequences, concept of video data mining is defined in a narrow sense and a broad sense respectively. The broad definition is classified into two categories: descriptive video data mining and predictive video data mining. This paper proposes a technique taxonomy framework, discusses some existing approache. It makes an expectation on valuable future work in this area.

Key words: video data mining, descriptive video data mining, predictive video data mining

中图分类号: 

• TP393