[1]张 亮,朱振峰,赵 耀,等.基于镜头的鲁棒视频广告检测[J].智能系统学报,2007,2(02):83-88.
 ZHANG Liang,ZHU Zhen-feng,ZHAO Yao,et al.Video commercial detection based on the robustness of sho t[J].CAAI Transactions on Intelligent Systems,2007,2(02):83-88.
点击复制

基于镜头的鲁棒视频广告检测(/HTML)
分享到:

《智能系统学报》[ISSN:1673-4785/CN:23-1538/TP]

卷:
第2卷
期数:
2007年02期
页码:
83-88
栏目:
出版日期:
2007-04-25

文章信息/Info

Title:
Video commercial detection based on the robustness of sho t
文章编号:
1673-4785(2007)02-0083-06
作者:
张 亮1朱振峰1赵 耀1卢汉清2
1.北京交通大学信息科学研究所,北京100044;
2.中国科学院自动化所模式识别国家重点实验室,北京100080
Author(s):
ZHANG Liang1 ZHU Zhen-feng1 ZHAO Yao1 LU Han-qing2
1. Institute of Information Science, Beijing Jiaotong Univesity, Beijing 100044 ,China;
2.Key Laboratory of Pattern Identification, Institute of Automation under the C hinese Academy of Sciences, Beijing 100080, China
关键词:
视频广告检测镜头检测支持向量机
Keywords:
video commercial detection shot detection support v ector machine
分类号:
TP391
文献标志码:
A
摘要:
随着多媒体技术的发展,自动检测出数字视频节目里面嵌入的广告是很具挑战性的研究.然而,由于嵌入的广告的制作方式和表现手法的多样性,很多自动检测模型的实验结果往往不甚理想.为了提高检测系统的鲁棒性,提出了3阶段广告检测系统.首先,提出了基于区域特征重要性的镜头检测算法(RBFID, regionbased feature importance detect ion),实现视频播放中突变镜头和消隐镜头的检测, 同时从每个镜头提取出一些统计特征用来标识镜头.然后,利用SVM的优异分类特性实现镜头分类.最后为了能得到精确的广告视频段,利用广告视频在内容和时间上的连续性来消除错分的镜头,然后将广告镜头整合成广告视频段.本系统在30个电视节目的片段上进行验证,实验结果表明此广告检测系统具有实用性.
Abstract:
Automatic detection of commercials embedded in digital video materials is a challenging task with the development of retrieval of multimedia data. How ever, because of the diversity of production modes and expression methods of com mercials, the performances of current detection systems are inadequate. In this paper, a threephase system for commercial detection is suggested to improve ro b ustness of the detection system. Firstly, two Regionbased Feature Importance D e tection schemes are proposed to detect cut shots and dissolved shots respectivel y and some statistical features are also extracted to mark shots. Secondly, an S VM classifier is applied to classify these shots. Finally, in order to obtain co mmercial segments more accurately, a statistical comparison of time and content of commercials is used to eliminate falsely cut shots. Finally, these commercial shots are integrated into commercial video segment. Test results on 30 TV video segments show the effectiveness of the suggested system.

参考文献/References:

[1] THOMAS HARGROVE url: http:// thomashargrove.com/logo detection/,2005-08-28.
[2]HAUPTMANN A, WITBROCKM, Story segmentation and detection of commercials in broadcast news video [A]. Advances in Digital Libraries Conference [C]. San ta Barbara, CA, 1998.
[3]DAVID A. SADLIER. Automatic TV advertisement detection from MPEG bitstream [ A]. International Conference on Enterprise Information Systems [C]. Setubal, Portugal,2001.
[4]HUA Xiansheng, LU Lie, ZHANG Hongjiang. Robust learningbased TV c ommercia l detection [A]. International Conference on Multimedia and Expo (ICME’05) [ C]. Amsterdam, Netherlands, 2005.
[5]PINAR DUYGULU. Comparison and combination of two novel commercial det ection Methods [A]. International Conference on Multimedia and Expo (ICME’04) [C]. Taipei, China, 2004.
 [6]RAINER L. On the detection and recognition of television com me rcials [A]. Proc of IEEE Int’l Conf. On Multimedia Computing and Systems [C] . Ottawa,Canada,1997.
 [7]MIENE A,DAMMEYER A,HERMES T H, et al. Advanced and adaptive shot bounda ry detection[J]. Proc. of ECDL, 2001(8)39-43.
[8]VAPNIK V. The Nature of Statistical Learning Theory[M]. New York Springer , 1995.

备注/Memo

备注/Memo:
收稿日期:2006-10-26.
基金项目:
国家自然科学基金资助项目(60373028、90604032、60602030) ;
教育部博士点专项基金资助项目(20030004016);
新世纪优秀人才支持计划及NLPR国家重点实验室开放基金资助项目.
作者简介:
张 亮,男,1981年生,硕士研究生,主要研究方向为基于内容的视频检索.
 E-mail:zhangliangwin@hotmail.com.
 朱振峰,男,1974年生,讲师,主要研究方向为目标检测与识别、图像检索、视频分析以及机器学习. 
赵 耀,男,1967年生,教授,博士生导师,北京交通大学特聘教授,北京交通大学信息科学研究所所长,2004年入选教育部新世纪优秀人才[JP2]支持计划.主持和参加了24项科研课题,发表论文90余篇,目前50余篇被SCI、EI检索,出版著作2部,申请发明专利3项.
更新日期/Last Update: 2009-05-06