Research Of Video Fingerprint Extraction Method For Content Security Monitoring

With the development of Internet technology, computer technology and multimedia technology, meanwhile smart mobile devices are becoming increasingly popular, network video traffic data has been tremendous growth. Although the network video services has brought great convenience to people’s daily lives, but at the same time unsafe, unhealthy video content also brought great harm and adversely affected to society. Because of the open nature of the Internet and with the video editing software features innovative and rich, so that people can more easily access, edit and re-publish video content and and other operations, How effective monitoring video content safe has become a new network monitoring difficulties. Content-based video fingerprinting as a potential and highly innovative techniques, has attracted more and more enterprise and security research institutions’ s attention and research.Video fingerprints should generally meet the robustness, accuracy and some real-time. This article describes the existing content-based video fingerprint extraction algorithm and analyzes their existence problems and shortcomings. To further study the content-based video fingerprinting algorithm, the main work is as follow:Firstly, analysis of the current network video business faced with the related questions of the content security monitoring, explanation the background and significance of the paper content’s study. And introduces the present status and study of network video content security monitoring at home and abroad.Then, introduced the mechanisms and technologies for video content security monitoring, respectively discusses digital watermarking and digital fingerprinting of the information hiding technology in the application of digital content protection, mainly introduces the fingerprint extraction algorithm based on video content and analyzed their performance.Then, the basic concept of SURF algorithm are introduced. In order to reduce the time cost of fingerprint extraction process and enhance operational efficiency of K-means clustering algorithm, SURF algorithm has been improved in this paper. On the one hand, using the form of concentric circles feature descriptor computation window, not only reduces the descriptor feature vector dimensions, while also eliminating the direction of the feature point distribution operations. On the other hand by adding diagonal Haar template strengthen the expression of feature descriptor ability. And through the Matlab simulation platform verifies the SURF of the improved algorithm has good accuracy and robustness.Finally, this paper introduces the K-means clustering algorithm and the concept of visual vocabulary, then video fingerprint generation method is proposed on the basis of front. Because the fingerprint generation using the local invariant features of video, ensure the accuracy and robustness of the fingerprint. While introducing the concept of visual vocabulary, make easy to store and retrieval video fingerprint. And through the Matlab simulation platform verification algorithm is proposed in this paper, from the result, we know that this method has good robustness and accuracy, and can meet certain real-time.