计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (22): 184-188.

• 图形、图像、模式识别 • 上一篇    下一篇

音视频相结合的广告检测算法

丁汝一,杨  宁,董道国   

  1. 复旦大学 计算机科学与技术学院,上海 200000
  • 出版日期:2012-08-01 发布日期:2012-08-06

Robust video and audio based commercials detection

DING Ruyi, YANG Ning, DONG Daoguo   

  1. Department of Computer Science, Fudan University, Shanghai 200000, China
  • Online:2012-08-01 Published:2012-08-06

摘要: 利用广告最后一帧的特点,提出了一种新颖的音视频相结合的广告检测的方法。在镜头分割的基础上,运用基于SVM的FMPI(有产品信息的图像帧)检测与静音段检测相结合的方法检测广告区域最后一帧。根据检测出来的结果确定广告的开始位置和结束位置,从而确定广告区域。实验结果表明,该方法不仅具有较高的准确率和查全率,而且解决了以往广告检测边缘定位不准的问题。

关键词: 有产品信息的图像帧(FMPI), 广告最后一帧(EFC), 支持向量机分类器, 广告检测, 静音段检测

Abstract: Taking advantage of the feature on EFC(the Ending Frame in the Commercial), a new method of commercial detection is proposed in this paper. This method is based on shots, and applies FMPI(Frames Marked with Product Information) detection using SVM and silence detection to find EFC. According to the EFCs, the method predicts the beginning and ending frames in commercials and then gets the commercials’ boundary. Experiments show that the proposed method is not only able to obtain high detection accuracy and recall, but also resolves the problem in how to identify the boundary of the commercial block.

Key words: Frames Marked with Product Information(FMPI), Ending Frame in the Commercial(EFC), Support Vector Machine(SVM) classifier, commercials detection, silence detection