基于内容的视频拷贝检测的研究.pdf

太原理工大学硕士研究生学位论文视频检索的准确率;然后,提取关键帧的SIFT特征,并且利用著名的球哈希算法对SIFT特征建立索引,把128维的高维浮点向量变为二值索引,提高了检索速度。实验结果表明该方法不仅提高了视频检索的准确率和检索速度,而且保证了较高的查全率。实验证明,与基于灰度序方法、基于GIST特征的方法、基于SIFT特征的方法和基于ORB特征的视频拷贝检测方法相比,本文提出的方法在保证较高查全率的同时,有较好的检测精度,同时检索时间也得到极大提升。最后,利用本文研究的内容采用VC技术设计并实现了基于球哈希和改进的灰度差算法的视频拷贝检测系统,该系统具有较强可移植性和可扩展性。关键词:版权保护,视频拷贝检测,灰度差,SIFT特征,球哈希 II 万方数据太原理工大学硕士研究生学位论文ⅥDEO COPY DETECTl0N RESEARCH BASED 0lNTHE CONTENT AB STRACT The popularity oftheIntemet,a variety pression technology and high—capacity storagetechnology make thespread,production,storage ofthe video more growth inthenumber ofvideo generates alotof approximation repeat largenumber ofnearduplicate videomake the copyright protection ofthevideo very difficult,SO thevideo copy detection technology emerges. In recent years,many scholars engage inthestudy ofvideo copyright aretwo main methods ofvideo copy detection technology,one isbased on content and the other isbased on digital watermarking digitalwatermarking technology,the videosneed tobeembedded thewatermark protection beforethevideoreleased,embedded digitalwatermark inthe video Caneasily bedestroyed and losecopyright protection,SO thedigitalwatermark tostillexistsome limitationsinpractice. The method based on content withoutany embedded information intothevideo, Candirectly extract robustfeaturesfromvideo content itself,SO itisconvenient and detection process isdivided into the following fourmain steps:videopreprocessing,creating hashing index,feature III 万方数据太原理工大学硕士研究生学位论文 extractionandfeaturematching. Thekeyproblem ofvideocopyright protection is notonly thelowspeed of retrieval which isleaded by alarge amount ofvideo data,but alsothelow accuracy ofretrievalwhich iscausedby video crop,rotate,convert othervideo encoding format copy ,looking forgood robustness and high dimensional indexing algorithm e the focuses of video copyright protection. Inthispaper,a video copy de