Research On A Porn Video Detecting Algorithm Based On Motion Features

The 21st century is an era of highly information, with the rapid development of Internet technology, more and more information transmitting is to rely on networks. Internet pornography, violence and other harmful information has also aroused widespread concern in society, the community, especially teachers, parents of students required to take effective measures for the growth of young people to creat a good network of cultural environment. In order to protect national network of cultural security, this paper presents the detecting algorithm for porn video based on motion features.The algorithm is proposed based on the motion vectors from mpeg compressed video stream. Since the motion vectors reflected the real movement of the human in the video. By extracting the motion vector of the P and B frames from mpeg streams, the magnitude and direction of the features is extracted respectively by analyzing the human law of the motion. Then the detection algorithm proposed in this paper is used for testing. Experimental results demonstrate that the correct recognition rate of this algorithm is up to 80%, so it is fit for the most of video detection environments.In recent years, the use of HMM to classify various human actions have been rather broad. By initializing the HMM, extracting the three motion features of the video frame sequence,signal distribution,then The feature vectors are fed to Hidden Markov Model for training and classification of actions. Two actions were trained with distinct HMM for classification. Experiment results show that the correct recognition rate is 90%.