An Objectionable Video Detection System Based On Energy Curves In Multiple Color Space

With the rapid development of Internet and multimedia technology, many objectionable videos swoop when people watch video programs through the Internet, these videos often contain scenes of violence, gory, horror and pornography and so on, which are seriously harmful to physical and mental health of young people, and become an important factor that may lead to juvenile crime. In order to control the spreading of the bad video and create a good internet environment, it has become the focus of social attention to how to control the spreading of bad video. It is also a challenge to research the bad video detection.Based on previous work in our research group, this paper has designed and implemented objectionable videos detection system that based on energy curves in multiple color space. The system is designed to detect objectionable video clips in the screen, so as to control effect on objectionable video when playing. In this paper we match video features playback with the video feature in database updated constantly because of a large amount of information of many kinds of videos. When users find that there are bad segments in the video, it is able to mark the fragments and upload to the video feature library for users. This system contains two parts: client and server. The main functions of the two ends are as follows:The main function of Client is that using wavelet analysis technique to extract energy curves features to match with local video feature database when playing video, if matched, the objectionable video clips will be shielding. If there are bad video clips that do not block processing during the play, the user can tag the video and upload energy curve characteristics of video fragment and video information to the server side, after confirmed by the server end, video feature and video information will be added in video feature library. The client can regularly update the local video feature library.Server can show the tag information that the client upload, and confirm the bad video clips according to the uploaded information. We can obtain the relevant information and energy curve characteristics of the video when playing the selected video clips. If this video is bad video footage, the energy curve characteristics and video information will be saved to video features database; if it is not, this information will be canceled, and staffs of server will confirm the next information uploaded.The system can detect and shielding bad video clips in the video playing, and it can upload information marked to the server. But the system is made for local video playing. At present, the video format it can play is only AVI format, video playback and matching algorithms of video feature library still need to be improved, these are the important content to continue to study in the future.