Research On Intelligent Overlay Video Advertising

With the fast growth of internet in the last over 10 years and the rapid adoption of digital capture devices, more and more home and professional users begin to record and share their daily life instance, great event reports and so on. The quantity of online video and online video sharing website is increasing dramatically. As the main profit-making way for online services, online video advertising has become research hotspots at home and abroad.This thesis presents a contextual video advertising system, called AdOn, which supports intelligent overlay in-video advertising. Unlike most current ad-networks such as Youtube that overlay the ads at fixed locations in the videos (e.g., on the bottom one fifth of videos 15 seconds in), AdOn is able to automatically detect a set of spatio-temporal non-intrusive locations and associate the contextually relevant ads with these locations.The major results in the dissertation are as follows:The information structure of the video is obtained by video analysis technique, and the highlight shots will be used as the advertising timestamp which will maximize the effectiveness of advertising. Furthermore, the overlay ad locations are obtained on the basis of face and text detection, as well as visual saliency analysis, so that the intrusiveness to the users can be minimized. Meanwhile, the ads for specific video shots are selected according to content-based multimodal relevance so that the relevance can be maximized.The innovative results in this dissertation are as follows:At first, we present an innovative contextual overlay in-video advertising solution driven by video content rather than only by the descriptive text for the user-marked video content. The contextual overlay in-video advertising process in this system is further formulated as a mathematical optimization problem to jointly maximize the relevance and minimize the intrusiveness. Secondly, we propose to automatically detect the overlay ads locations by spatio-temporal analysis of video content. At last, we propose to match the ads with the specific video segments for overlaying the ads rather than the whole video sequence. In this way, the ads would be more contextually relevant to the ongoing video content.AdOn represents one of the first attempts towards contextual overlay video advertising by leveraging information retrieval and multimedia content analysis techniques. The experiments conducted on a video database with more than 100 video programs and 7,000 ad products indicated that the proposed solution is superior to existing advertising approaches in terms of ad relevance and user experience.