基于位置敏感哈希的网络视频重复检测  

Near-duplicate Web video detection based on locality sensitive hashing

作　　者：王洪峰[1] 刘辛[1] 

机构地区：[1]周口师范学院计算机科学与技术学院,河南周口466001

出　　处：《计算机应用研究》2012年第5期1954-1958,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61103143)

摘　　要：针对当前网络上存在着大量的重复或近似重复的视频问题,提出了一种基于镜头层比较和位置敏感哈希的快速准确的网络视频重复检测方法。通过视频间匹配的镜头数占查询视频总镜头数的比例来判断视频的相似性。除此之外,还利用著名的近似最近邻查找技术——LSH在镜头层来快速查找相似镜头,从而提高检测速度。通过将镜头作为检索单元,把数据库中所有视频的镜头放到一起构建一个新的数据集,将种子(查询)视频的每一个镜头作为一个查询请求,应用基于LSH的近似近邻检索方法,检索出与查询镜头相匹配的所有镜头,最后融合这些返回的结果,得到查询视频的重复或者近似重复的视频集。通过在包含12 790个视频的CC_WEB_VIDEO数据集上的实验结果表明,该方法取得了相比已有方法更好的检测性能。The amount of video clips on the Web is huge and is still growing very rapidly. However, there are many duplicate or near-duplicate videos on the Web, which affects the efficiency of video retrieval. This paper proposed a fast method to de- tect near-duplicate videos based on shot level comparison （SLC） and locality sensitive hashing （LSH）. It denoted the videos similarity by the ratio of the number of matched shots between two videos, and introduced the LSH in a special way to improve the speed of the detection. That was, instead of calculating the similarity between seed video and all others one by one, it tried to find all shots which were approximately matched with those in query video by LSH technology, and a rearrange dataset. Then, it could detect the near-duplicate videos of query, video by fusion the returned shot-sets. It applied this method on the CC WEB_VIDEO dataset include 12 790 videos. The experimental results demonstrate superior performance of this method o- ver other popular state-of-the-arts methods.

关 键 词：网络视频 重复检测 位置敏感哈希 镜头层比较 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]