序列模式挖掘在视频分类中的应用  

Application of Sequence Pattern Mining in Video Classification

作　　者：曹烈[1] 沈钧毅[1] 原野[1] 

机构地区：[1]西安交通大学计算机系,陕西西安710049

出　　处：《微电子学与计算机》2004年第3期73-75,112,共4页Microelectronics & Computer

基　　金：国家自然科学基金资助项目(60173058)

摘　　要：视频流的数据量大,又是一种非结构性的数据,因此视频分类一直是视频分析工作中的一个难点。本文采用视频分割技术,将视频分割成一系列的镜头,利用这些镜头在时间上的序列特征,使用序列模式对镜头序列进行分析,并提出一个基于序列模式挖掘的视频分类系统模型。该系统中主要利用cSPADE算法来实现对镜头序列的分类,以形成分类规则库,通过该规则库匹配待分类视频的镜头序列,从而实现视频的分类。Video stream has a great number of data and the video data is non-structured data,so video classification has been a difficulty in the field of video analysis for a long time. This paper uses the technology of video segmentation to divide the video into a series of shots and because these shots have sequential characteristic along the time,we can analyze the shot sequence using sequential pattern mining. A video classification system model based on sequential pattern is also proposed.This system mainly uses cSPADE algorithm to dispose of the video classification and forms a classification rule-lib to match the shot sequence that will be classified. Through this ,we can realize the classification of video.

关 键 词：数据挖掘 序列模式挖掘 视频分类 视频流 数据库 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]