The Feature Extraction And Recognition Methods Of Television And Radio Figure Images On Triple Play

With the coming of the era of triple play, the free dissemination of massiveinformation and the opening source information make the information more difficult andsteep to supervise. It became the brunt of problem to be solved which to supervise avariety of media (text, images, sound, video) content under the triple play and to controlthe generation and dissemination of the pornography, violence, anti-governmentinformation. In order to monitor these massive unstructured data,"content regulationsystem under triple play research project was born.HUSTRIM-PM, the picture monitoring system, determined its own role andassociated with each other systems which were in parent-HUSTRIM goals, functions,workflow and architecture.HUSTRIM-PM detailed the characteristics of the character image filteringtechnology from the color, texture angle, gave skin color filtering technology based onHSI color space and R values, multi-resolution analysis based on wavelet transform andtexture filtering technique based on the fractal dimension, used "fencing algorithm" toextract suspicious areas of sensitive pornographic images. in-depth studied of artificialneural network technology, selected the character image feature recognition technologybased on kohonen and BP neural network, achieved to identification Image sensitive areaUnder the interception of image feature extraction module.The architecture and the database of the image monitoring system has been designed,through simulation test sample, the effectiveness of these methods in the imagemonitoring system has been reflected. HUSTRIM-PM meets the design demand ofHUSTRIM system on the image monitoring aspects, with the features of real-time, highrecognition rate, makes some exploration in the area of character image featurerecognition.