摘要
针对视频拷贝检测中检索速度问题,提出一种基于关键帧多特征融合的类局部敏感哈希索引方法,将存在拷贝片段的视频映射到同一个哈希桶中,减少检索的范围,达到提高检索速度的目的。该算法首先对视频进行镜头分割提取关键帧,为了提高检测精度,分别提取了灰度序全局特征、基于关键点的希尔伯特特征、ORB(Oriented FAST and Rotated BRIEF)局部特征,综合利用全局特征和局部特征两者各自的优势;然后根据视频关键帧序列建立了类局部敏感哈希索引,利用建立好的索引获得拷贝检测结果。实验结果表明,该方法在保证检测精度的同时,速度上也有很大提升,具有重要的应用价值。
Retrieval speed is an important issue in video copy detection .This paper proposed a fast video copy detection method ,w hich uses a local sensitive hashing index to achive fast retriev‐al by mapping videos with similar clips into the same buckets ,and combines multi‐features of key frames to achieve high accuracy .To improve retrieval accuracy ,the method extracts key frames by using shot segmentation techniques ,and then obtains the Hilbert feature based on key points , ordinal measure feature and ORB （Oriented FAST and Rotated BRIEF ） feature from key frames . This allows the method make full use of both local features and global features .The hash index is built by hash the combined features .Experimental results show that this proposed method not only achieves high precision and recall rate ,but also has high spead .
出处
《太原理工大学学报》
CAS
北大核心
2015年第5期566-570,共5页
Journal of Taiyuan University of Technology
基金
国家科技支撑项目子课题:(2012BAH04F02-2)
教育部归国留学人员科研启动基金(2010-1174)
人社部留学人员科技活动项目(2011-508)
山西省国际科技合作计划基金资助项目:中美
中加合作云计算环境下网站系统架构和数据库系统建模技术研究(2012081034-2)
关键词
视频拷贝检测
关键帧
ORB
类局部敏感哈希
关键点希尔伯特
video copy detection
key-frames
oriented FAST and Rotated BRIEF （ORB）
local sensitive hashing
hilbert feature based on key points
