The Method Of Extracting Keyframes For Video Content Under The Triple-play Service

With the development of triple-play service, it has gained more and more nationalpolicy support. More and more operators eager to join in the movement and more andmore people love it. The real triple-play world is bound to come eventually. What isbrought by triple-play service is the free dissemination of mass information. The directresult of open source information is the increased difficulty of monitoring the informationcontent of the infinite. And after the triple play, network would be one of the daily needsof programming. It will have a major impact on people’s production, life, and ideas. Afterthe triple play, monitoring content of the network and control the generation anddissemination of ill information involving violence, anti-government has become anurgent need. However, mature technology of supervising such a large scale ofunstructured data has not yet appeared. And variety of monitoring and control techniquesof big data is just beginning or has yet to hit the road.Thus, the research projectbroadcasting content monitoring system for triple-play service came.As video content key frame extraction system’s parent system, HUSTRIM system isintroduced from three aspects, the objectives, functions and architecture, to make clearthat where the video content key frame extraction system is in HUSTRIM system andinterrelationship with other components in HUSTRIM system. The video contentidentification subsystem is re-combed and improved. According to the objectives and therequirements of HUSTRIM system, the specifically definition of video key frames inHUSTRIM system is given.The method in content-based video retrieval field isintroducted in detail for deep learning of video key frame extraction technology base onuncompressed. And according to the definition of HUSTRIM video key frames, a newmethod of video key frame extraction techniques, based on fuzzy shot and anglicizing themovement of foreground, is proposed, by transforming the shot segmentation based on YUV Eigen values is,and combined with motion analysis technology of front object. Andthe video content key frame extraction system, subsystem of video content identificationsubsystem, is designed and implemented by using the new method proposed above, withthe assist of OpenCV, an open source library. And comprehensive tests have been given tothe system in comparison.The simulation experiments improve that the system is better than the last one onfunctions and is faster as well. However, it needs to be further improved for mass videoinformation. Some suggestion for further improvement is given at last according to theformer work.