Research And Implementation Of An Objectionable Image Detection System Based On Skin Color And Facial Information

The network culture with Internet developing rapidly has been a new trend of the current cultural development. Internet easy access brings an unprecedented pace and breadth in both time and space for the spreading and the carry-forwards of Internet culture. The combination of the culture and the network has given the social culture new vitality, enriched the human spirit and brought our lives with far-reaching impact. But at the same time, the low regulatory of the network dramatically increased the risk of cultural healthy developing. More and more network rovers are suffering from the harassment or threaten from much objectionalbe information. On this issue the most concern is the potential impact and the solutions of the objectionable information in the Internet culture for the young people who lack of sense. And currently it has become a major worldwide topic.In this thesis, we have a study on the detection and analysis of the pornographic information involved in the images of the network combined with the background of the network culture security research. After understanding of the current objectionalbe image detection technology, the thesis made a deeply research on the basic image color attribute, the concept of different color space, color space conversion, skin detection and skin segmentation technology, face detection technology and target detection classifier technology, then designed an objectionable image detection system based on the skin color and facial information, and devoted to support the development of the network cultural security project as underlying technologies.The thesis introduces the structured framework of system designation and the system process of the development in detail, including the different kind of images reading process module, the attributes obtaining and type analyzing module of the image objects, the image color space transformation module, skin detection, skin segmentation and statistics module, the face detection module as well as systems integration and other modules. It achieves the analysis and testing of the objectionable characteristics for any network images through the development of the system. The experiments show that the system is not only consistent with the principle of simplified development but also achieves good detection results.