Research On Porn Image Filtering And System Implementation Based On SVM

With the rapid development of computer technology and networktechnology, data communication and resource sharing become the mainfunction of the computer network. The probability of people touchingunhealthy information is increasing. This information will have anegative impact on internet users especially youth people. Filteringunhealthy information in the client is an effective solution, and is theresearch contents in this paper.Firstly, the images containing unhealthy information are analyzed.The processing technology and implementation method about images aredescribed. The method which obtains three aspect eigenvalues of imagecolor, texture and geometry is proposed. The porn images areidentificated by using SVM method.Secondly, a general framework of filtering porn image is proposedbased on image eigenvalue, and the weights strategy identificationmethod is joined. After obtaining the eigenvalue, normal images isidentificated by using quickly determine method. Then combination oftwo methods to achieve the classification, one is the method based on theweights strategy of each eigenvalue, another is the eigenvalues as a wholevector classification, and experiments show that the method has betterrecognition accuracy and time efficiency.Finally, the unhealthy information filtering system has beenimplemented based on SVM, web data is analyzed in real time, and thecurrent web page will be identificated through image and text analyses.Many of the unhealthy information web pages are shielded in realenvironment and meet the requirements, contribute to the networkenvironment maintain healthy.