Research Of Identification The Key Parts Of Sensitive Images

Filter the sensitive images on the network is based on identification of sensitive images, many of sensitive image filtering method is based on the text which attached to images or URL address information, these filtering technology is lack of the dynamic and real-time to identify the sensitive images, thus Content-based image recognition has become a sensitive issue to be resolved urgently.The traditional content-based sensitive image identification primarily on the basis of the image area with the body color of the characteristics to classification, this method is relatively simple and quick, but it is often with a higher false rate. In order to lower the false rate of the traditional content-based sensitive image identification, we conducted a study of content-based sensitive image identification, the research is based on the human color segmentation, and divided into three parts, these parts are face recognition, human body posture recognition and human key parts identify. This paper is a stage summary of the human key parts identify of sensitive image. The research goal is to locate and identification the human key parts of sensitive image.The main research include the following:1. Improve the FCM clustering algorithm:1.1 establishment of hierarchical image cone. To reduce the size of the image data to reduce the image clustering segmentation computation.1.2 introduction of spatial information. Use image normal pixels information shielding the noise pixel information, enhance the anti-noise performance.2. Improve the human color segmentation methods:2.1 Color model segmentation. Establish the Gaussian color model on CbCr plane to segment human color area.2.2 Color clustering judgments. Judgment the clustering of the color to improve the accuracy of color segmentation.3. Identify the key parts of body:3.1 Identification the key parts of female upper body. The brightness of color through the image location identification information.3.2 Identification the key parts of female lower body. Filter through the eight square windows identify matching position.3.3 Identification the key parts of male lower body. Through the edge of the shape of image segmentation similar to the level of sentences.