基于运动矢量统计的不良视频检测算法研究
|
摘要
随着网络科技的日益普及与深入发展,Internet已经成为大众生活的重要组成部分。互联网在给我们带来便捷的同时,也带来了安全隐患,其中渲染黄色、暴力等不良内容的文字、图片、网络视频和网络游戏更是成为茶害青少年健康成长的‘信息罂粟’。如何营造绿色健康的网络环境、纯净虚拟的网络空间,切断网络蔓延开来的无形毒手,成为广大家长、师生的共同呼声。针对网络中的不良视频,本文提出了一种简单而快捷的检测算法,设计了相应的检测过滤系统。
本文提出的检测算法主要是基于H.264压缩视频中的运动矢量统计与归一化计算(本文提出的归一化,是指一视频段内,帧间宏块运动矢量换算成相对与固定参考帧的位移偏移量)。对提取的运动矢量场先去噪滤波与内插加强,进行归一化计算,得一稳定的主运动矢量场值,然后观测组帧图GOP内的∑MV数列特征时域曲线波形,或匹配每一组帧间的时序曲线特征,得出镜头主要运动目标的运动轨迹与特点,当运动目标呈现人体敏感部位运动特征时,则作出判断,从而快速对网络数据流进行阻断选择。
Along with the network technology popularization and development, Internet has become an important part of public life.When the Internet brings convenience to us, also bring safe hidden trouble. those been naked describe of erotic, violence and other undesirable content online video and online games are poisoning people brainwashed minds and becoming the healthy growth of teenagers'information Poppy', how to create a healthy network environment of green, clean virtual network space, cut off the network'spread and invisible pernicious hands', become the common voice of the majority of parents、teachers and students.Aiming at the erotic video network, this paper presents a simple and fast detection algorithm, designed the corresponding detecting and filtering system.
The proposed detection algorithm is calculated and normalized H.264compressed video based on motion vector.The motion vectors extracted firstly normalized calculation, and then to the dislodge hash filter and interpolation to strengthen, the motion vector field values to robust a normalized,∑MV sequence characteristics and comparison of time domain on each frame set, the trajectory and characteristics of the motion target, when the target is sensitive parts of the body motion characteristics, then to judge, and select pass or blocking-up the bit stream.
本文提出的检测算法主要是基于H.264压缩视频中的运动矢量统计与归一化计算(本文提出的归一化,是指一视频段内,帧间宏块运动矢量换算成相对与固定参考帧的位移偏移量)。对提取的运动矢量场先去噪滤波与内插加强,进行归一化计算,得一稳定的主运动矢量场值,然后观测组帧图GOP内的∑MV数列特征时域曲线波形,或匹配每一组帧间的时序曲线特征,得出镜头主要运动目标的运动轨迹与特点,当运动目标呈现人体敏感部位运动特征时,则作出判断,从而快速对网络数据流进行阻断选择。
Along with the network technology popularization and development, Internet has become an important part of public life.When the Internet brings convenience to us, also bring safe hidden trouble. those been naked describe of erotic, violence and other undesirable content online video and online games are poisoning people brainwashed minds and becoming the healthy growth of teenagers'information Poppy', how to create a healthy network environment of green, clean virtual network space, cut off the network'spread and invisible pernicious hands', become the common voice of the majority of parents、teachers and students.Aiming at the erotic video network, this paper presents a simple and fast detection algorithm, designed the corresponding detecting and filtering system.
The proposed detection algorithm is calculated and normalized H.264compressed video based on motion vector.The motion vectors extracted firstly normalized calculation, and then to the dislodge hash filter and interpolation to strengthen, the motion vector field values to robust a normalized,∑MV sequence characteristics and comparison of time domain on each frame set, the trajectory and characteristics of the motion target, when the target is sensitive parts of the body motion characteristics, then to judge, and select pass or blocking-up the bit stream.
引文
[1]http://www.techweb.com.cn/news/2009-06-15/410937.shtml.
[2]http://news.sohu.com/20090609/n264434376.shtml.
[3]http://news.163.com/09/0609/14/5BCGT3NQ000120GU.html.
[4]http://news.xinhuanet.com/zhengfu/2004-07/19/content_1612816.htm.
[5]吕丽,杨书堂等.基于光流法的不良视频检测算法.计算机工程,2007,33(12).
[7]蔡群,陆松年等.基于音视特征的视频内容检测方法.计算机工程,2007,(22).
[8]雷斌,郅琦.基于MPEG运动矢量的行车安全检测算法.西安工业大学学报,2007,(1)
[9]陈杉,于鸿洋.基于MPEG压缩域的运动对象检测方法.信号处理,2004,20(6).
[10]刘艳民,安红心.基于运动特征的不良视频检测算法.微计算机应用,2010,(7).
[11]Forsyth D A, Fleck M M. Identifying Nude Pictures[C]//Proc.of the 3rd IEEE Workshop on Applications of computer Vision.1996.
[12]Jeong C Y, Kim J S, Hong K S.Appearance-based Nude image recognition[c]//Proceedings of the 17th International Conference on Pattern Recognition.2004:1051-4651.
[13]陈攀.基于特征部位和肤色的不良视频检测的算法研究.华中师范大学,2011.
[14]雷兰一菲,郎海涛.几种典型局部图像特征的比较.计算机应用,2012,(S2)
[15]柳伯超.基于内容的不良图像识别研究.山东师范大学,2007.
[16]孙竞媛.基于内容的敏感图像过滤技术的研究.吉林大学,2007.
[17]侯冠华,史萍.视频分割与场景聚类算法研究.中国传煤大学学报：自然科学版,2006,13(2)
[18]朱映映,周洞汝.一种从压缩视频流中提取关键帧的方法.计算机工程与应用,2003.(18)。
[19]陈兴杰,许劲松.基于视频的入侵检测研究.微计算机信息,2010,(12).
[20]王丽.基于H264的运动对象快速分割算法.计算机工程,2011,(12).
[21]丁文锐,杨桦等.利用H.264中运动矢量实现运动目标的检测.北京航空航天大学学报2010,(02).
[22]颜建强,高新波等.多媒体信息服务中色情图像检测算法.中国体视学与图像分析,2011.(1).
[23]毕厚杰主编.新一代视频压缩编码标准——H.264/AVC.人民邮电出版社出版,2009.
[24]张智福.从H264压缩码流提取运动对象的一种方法.计算机工程,2010,(19).
[25]柴桦.基于视频的人体运动识别方法研究.中南大学,2008.
[26]李德禄.室内视频监控中行人目标检测与跟踪方法.福建电脑,2008,24(5).
[27]范志斌.基于图像局部不变性特征的机器人目标识别与定位.北京交通大学,2011.
[28]吴心筱.图像序列中人的姿态估计与动作识别.北京理工大,2010.
[2]http://news.sohu.com/20090609/n264434376.shtml.
[3]http://news.163.com/09/0609/14/5BCGT3NQ000120GU.html.
[4]http://news.xinhuanet.com/zhengfu/2004-07/19/content_1612816.htm.
[5]吕丽,杨书堂等.基于光流法的不良视频检测算法.计算机工程,2007,33(12).
[7]蔡群,陆松年等.基于音视特征的视频内容检测方法.计算机工程,2007,(22).
[8]雷斌,郅琦.基于MPEG运动矢量的行车安全检测算法.西安工业大学学报,2007,(1)
[9]陈杉,于鸿洋.基于MPEG压缩域的运动对象检测方法.信号处理,2004,20(6).
[10]刘艳民,安红心.基于运动特征的不良视频检测算法.微计算机应用,2010,(7).
[11]Forsyth D A, Fleck M M. Identifying Nude Pictures[C]//Proc.of the 3rd IEEE Workshop on Applications of computer Vision.1996.
[12]Jeong C Y, Kim J S, Hong K S.Appearance-based Nude image recognition[c]//Proceedings of the 17th International Conference on Pattern Recognition.2004:1051-4651.
[13]陈攀.基于特征部位和肤色的不良视频检测的算法研究.华中师范大学,2011.
[14]雷兰一菲,郎海涛.几种典型局部图像特征的比较.计算机应用,2012,(S2)
[15]柳伯超.基于内容的不良图像识别研究.山东师范大学,2007.
[16]孙竞媛.基于内容的敏感图像过滤技术的研究.吉林大学,2007.
[17]侯冠华,史萍.视频分割与场景聚类算法研究.中国传煤大学学报：自然科学版,2006,13(2)
[18]朱映映,周洞汝.一种从压缩视频流中提取关键帧的方法.计算机工程与应用,2003.(18)。
[19]陈兴杰,许劲松.基于视频的入侵检测研究.微计算机信息,2010,(12).
[20]王丽.基于H264的运动对象快速分割算法.计算机工程,2011,(12).
[21]丁文锐,杨桦等.利用H.264中运动矢量实现运动目标的检测.北京航空航天大学学报2010,(02).
[22]颜建强,高新波等.多媒体信息服务中色情图像检测算法.中国体视学与图像分析,2011.(1).
[23]毕厚杰主编.新一代视频压缩编码标准——H.264/AVC.人民邮电出版社出版,2009.
[24]张智福.从H264压缩码流提取运动对象的一种方法.计算机工程,2010,(19).
[25]柴桦.基于视频的人体运动识别方法研究.中南大学,2008.
[26]李德禄.室内视频监控中行人目标检测与跟踪方法.福建电脑,2008,24(5).
[27]范志斌.基于图像局部不变性特征的机器人目标识别与定位.北京交通大学,2011.
[28]吴心筱.图像序列中人的姿态估计与动作识别.北京理工大,2010.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |