新闻视频中广告片段精确定位方法研究

发布时间：2018-05-13 16:03

本文选题：广告片段定位 + 聚类　；参考：《中国图象图形学报》2009年07期

【摘要】：由于广告节目制作风格的多样性,新闻视频中广告片段的检测和定位是一个非常具有挑战性的问题。提出了一种新闻视频中广告片段定位的新方法。首先针对镜头切换和非镜头切换数目不平衡问题,设计了一种新的分类方法进行镜头检测,然后利用聚类分析粗略地标识广告块,最后通过分析相邻镜头的平均持续时间和镜头关键帧的视觉特征准确定位广告边界。实验结果表明,该方法具有较高的定位精度。
[Abstract]:Because of the diversity of the production styles of advertising programs, the detection and location of advertising segments in news videos is a very challenging problem. A new method of advertisement segment localization in news video is proposed. First of all, aiming at the imbalance between shot switching and non-shot switching, a new classification method is designed for shot detection, and then the advertising blocks are roughly identified by clustering analysis. Finally, by analyzing the average duration of the adjacent shots and the visual features of the lens key frames, the advertising boundaries are accurately located. The experimental results show that this method has high accuracy.
【作者单位】：河海大学计算机及信息工程学院;复旦大学计算机科学与工程系;
【基金】：国家自然科学基金项目(60673141)
【分类号】：TP391.41

【参考文献】