Reseach On Vedio Fingerprint Retrieval Algorithm Based On Squence Alignment

With the strengthening of the network infrastructure and broadband technology is becoming more and more mature, Network transmission speed in our country has been greatly improved. Digital multimedia video information explosively growth in network. As the video brings great convenience to people, there are many illegal videos and infringement videos spreading on the Internet. How to ensure the Internet video be legal and secure attracts more attention of the network security regulators and the owner of the video resources. Aiming at the application scene of network video security, this paper focuses on the research of fast video retrieval algorithm based on fingerprint technology. It has great significance for the implementation of the fast video retrieval and Internet video supervision. The main contents of this paper are as follows:First we discussed the problem of the video content security supervision on the Internet. And on the basis of analyzing and discussing the research status of the video fingerprint technology at home and abroad, we summarize the problems to be solved of current video fingerprint retrieval and matching technology.Then we introduce the video retrieval and matching technology, and Separately discussed framework of the content-based video retrieval technology and the video fingerprinting technology. We also introduce several common video fingerprint matching technique.Then in the view of the problem of the current video fingerprint matching has not a good balance of accuracy and real-time technology, this paper put forward two layer fast video retrieval scheme based on the coarse-grained and fine-grained. The scheme includes two important algorithm, the first one is the algorithm of coarse granularity of video retrieval based on inverted index, the second one is the algorithm based on the improved BLAST algorithm for video fingerprint fine-grained retrieval. Inverted index retrieval algorithm in the indexing stage using a trained visual vocabulary to conduct word frequency statistics of characteristic points of the key frame and extract keywords through the statistical results, inverted index table are established by the keywords. While on the retrieval phase we obtain keywords by the same procedure, then we can fast screen candidate video through the keywords information by the inverted index table. This can reduces the comparison times for the Subsequent fine-grained accurate retrieval and greatly improved inefficient situation of the sequence alignment. For the video fingerprint fine-grained retrieval, this paper put forwards a method that using bioinformatic sequence alignment technique—BLAST algorithm to complete retrieval. But BLAST algorithm itself has efficiency defects applying to video fingerprint sequence alignment. Therefore, in order to improve the efficiency of BLAST algorithm, and to make it better apply to video fingerprint quick alignment, we put forward pattern matching algorithm to improve the efficiency of the BLAST algorithm. Finally, we verify the feasibility and high efficiency of the proposed coarse-grained and fine-grained fast video retrieval scheme by experiment and simulation.At the end of this paper, we summed up the main research contents and analyzed the shortcomings of the algorithm proposed in this paper. On this basis, I put forward some suggestions and prospects for the future research direction of my own.