Research And Implementation Of Content-based Video Streaming Detection

With the rapid development of the network and streaming media technology, videosharing sites and the portals are more and more popular, these applications providepeople with convenience, at the same time, it also brings a variety of problems, such asreleased harmful, illegal and pirated video content through the network, and how todetect network advertising effects quickly. How to manage the content of the on-linemassive video effectively is an urgent problem, video content detection and itsapplication of research results have a pressing demand and big market prospects. Thispaper introduces the current technology and background of video content detection, anddevelops a real-time content-based video streaming detection system. Firstly, we dosome researches about the existing shot detection and key frame extraction algorithms,and the sub-block histogram is the best algorithm to online detection. Secondly, weextract local features which called SIFT from the key frame, and match with the data ofthe candidate set, local features have high-dimensional and huge characteristics, Tosolve the problem of low efficiency about SIFT feature matching, we use hierarchyindexing technology of the high-dimensional point set. We generate high-level abstractcharacteristics and establish indexes for them, then establish secondary index for localfeatures and establish mapping relationship between the abstract features and itscorresponding high-dimensional point set. Finally, we introduce the concreteimplementation of the detection system. Our extensive experiments study on a largedatabase and demonstrates that the sub-block histogram, local feature extractionalgorithm, and hierarchical indexing technology can improve the speed and accuracy ofonline video detection, and get a good test results.