Research And Implementation Of Dynamic Face Recognition Technology Based On Mobile Video

Nowadays, the video resources become more and more abundant and the video-related services and entertainment applications become increasingly diverse at the same time with praise and dependence of users. In most video resources, the characters are generally the subject for users. It will greatly improve the user experience dramatically with unexpected surprise to users, if the relevant information of characters can be accurately extracted from the video, and to do some special processing to generate specific effect. Thus, the dynamic face recognition becomes research focus in the many fields such as image processing, computer vision, pattern recognition, machine learning, and artificial intelligence. Therefore, this thesis focuses on the dynamic face recognition technology.After introduce the existing related face recognition technology briefly, an optimized mechanism based on the existing dynamic face recognition system called face consistence detection mechanism has been proposed, as the poor quality and the complexity of video, especially for the video shot on the mobile terminal, or even the defects of the face recognition algorithm itself. The optimized mechanism fully considers the information correlation between video sequences, and introduces the idea of grouping and voting to determine the consistence and validity for face detection data, and takes preliminary face clustering at the same tune.Then the thesis describes the design and the implement of people video summary based on the improved dynamic face recognition system in detail. Together with a large number of comparative experiments and performance analysis, it proves the validity and usability of the face consistence detection mechanism. And the experimental results show that the face consistence detection mechanism significantly improves the dynamic detection and face recognition performance.Finally, the thesis summarizes the full text and puts forward future research direction for the shortcomings in current work. Then, it makes a work summary that the authors have done during graduate.