Research On The Method Of Clothing Advertising Based On Video Protagonist Identification

In this paper, we present a new method for video advertising: based-on the video protagonist identification for clothing advertising. This framework that aims to embed protagonist-connect and content-relevant advertisements with a video stream is thoroughly studied. As we all know, with the popularity of the Internet and mobile handheld devices, network video advertising has brought great benefits that has become a major source of income for many operators. But the advertising bundled video, forced consumers to watch it that has nothing to do with the video's content and advertisement playing time more and more long, which beyond the endurance of the audience and serious impact on the advertising effect. In order to reduce the intrusion of advertising on video and make the user experience don't get hurt, we present this new method for video advertising: based-on the character for video advertising in this paper.This paper completed the following work: First, we Analyzes the present situation in China and broad of the network video advertising, points out the problems of existing way of advertising. Thus, we proposed a new video advertising method: based-on the character for video advertising. The method enables ads that inserted related to the video content and connected to the protagonist of video, thus it can improve the audience's interest in advertising, meanwhile it also can increased the choice of the ways of video advertising. Second, we summary and analyzes the present situation in China and abroad of video face clustering technique. According to the characteristics of the video, we propose a clustering algorithm that based on face verification. The algorithm uses the transmission of face verification to achieve face clustering. In this paper, we use the technology of face detection, face verification, object tracking and so on, which improves the accuracy and stability of the video face clustering. Third, we analyze the present situation of the research on the retrieval and matching of clothes, and give a method for matching the clothes in video. We use the model to achieve the segmentation and classification of clothes, and the alignment of the clothes can be realized through the body posture alignment. Which can improve the accuracy of clothes matching. Finally, I achieve the advertising system based on video content. The system can detect many objects in the video, and able to match the clothing of video role, meanwhile it can recommend similar clothes to user.To sum up, this paper uses the method of face clustering to realize the automatic recognition of the protagonist in the video, and through the method of the recognition and retrieval of clothes, to realize the advertising on the main characters of the video, which in order to improve the user experience.