Using Webpage's Features Recognize Porn Webpage

In the present, whether in international or domestic people are generally considered to be toxic to young porn web page, endanger their physical and mental health, impede their healthy development, and some may even be forced into crime. Therefore to pornographic Web pages for a variety of information to distinguish the normal characteristics of information and bad information. The author combines research, the paper chose to study the Internet in the characteristics of pornographic images.Web page content for the site characteristics and size, supervision, we will contain the image of the website is divided into three categories of poor:the larger more popular forum, blog, they supervise the more stringent, small, or regulatory forum than the loose, blog, and contains a large number of pornographic images of porn sites. This choice of several larger communities, such as famous cat flutter and End of the World Community, a loose Chinese Microsoft Live community, as well as dozens of pornographic Web site for the study. Firstly, select a different website URL as a crawler crawling seeds, and then select a different crawling strategies, different depths of the web crawl. Extract pages of image data, and other web data, the data for further study.Based on the above data, we study include pornographic image content features, pornographic images from a large number of areas of skin exposed to this feature in mind. Such as:color, size and image area ratio, where the rectangular area with color, color, number of connected regions, the largest connected region and the image area ratio of the average variance of skin color, skin color, etc. The average probability, but also of the pornographic images with the surrounding media the relationship between content, such as pornographic images throughout the site accounted for the proportion of image content, the characteristics of pornographic text pages. But sometimes in the normal site or porn sites facial close-up when the protagonist, will be mistaken for pornographic images. Therefore, we further studied to distinguish the proportion of normal human face features.Although the normal site may also exist a lot of bad images, but because there is miscarriage of justice in the use of color characteristics of the situation, making detection of the normal site of images occurs when the number of false positives and opportunities for greatly increased. However, as we realize the statistics of the negative image of the frequency of view, the negative image most of the normal site are common high frequency images. We can study the characteristics of these images and judge, to get good results. Here we use the cluster through the SIFT features of the 128-dimensional images of the normal site to judge, so the normal site in the misjudgment of the situation greatly reduced.Finally, we give the experimental environment, complete these steps, using the experimental data shows the correctness of the theory test. Our experimental results are analyzed and shows that the advantages of each approach, and inadequacies. This method is given in different occasions of use proposals. Technical knowledge is to serve the people. Use of pornography is to determine the characteristics of technology, making use of camera camera, the monitor display and dissemination of pornographic images on the web is the use of multimedia technology. For young people, we must not only shield bad information, but we should understand their education, develop good outlook and outlook on life, give them a good family and social environment in order to make them on the road of life farther and better.