基于运动特征的不良视频检测算法  

Objectionable Video Detection Algorithm Based on Motion Features

作　　者：安红心[1] 刘艳民[1] 

机构地区：[1]兰州大学信息科学与工程学院,兰州730000

出　　处：《微计算机应用》2010年第7期32-37,共6页Microcomputer Applications

摘　　要：提出了一种适用于流媒体中的不良视频检测算法。首先从MPEG压缩视频流中提取运动矢量,然后对运动矢量进行预处理,剔除噪声干扰,从而提取出不良视频中的运动特征,通过设定阈值实现整段视频的判别。实验表明,该算法识别率达到80%,鲁棒性较好,适用于各类不良视频的过滤与检测。This paper proposed an objectionable video detecting algorithm which uses motion features from the videos. Firstly,it extracts the motion vector from compressed MPEG videos. Secondly,motion vector is pretreated to smoothing the interference,and the motion features are extracted from objectionable videos. Finally,with appropriate threshold,objectionable video can be detected. The experiment results show that this method may correctly detect objectionable video with a high-probability （up to 80%） and it is appropriated for the video detection environment.

关 键 词：运动矢量 运动特征 不良视频检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]