A Novel Classification Based On Sequential Pattern Mining In Videos

With the rapid advances in computer technology and multimedia techniques, multimedia information plays a more significant role in people's daily life in recent years. Multimedia information generally contains many forms, such as digital images, video, audio, dynamic graphics and text form. What's more, a great number of videos are adopted as a resource of information in our life for the reason that digital videos contain much more information than that of other forms.At the same time, with the worldwide popularity of internet applications, huge databases are constructed by digital videos. Therefore, it is much difficult to get the videos which you are interested from the database for users. Facing so huge quantity of video data, how to manage, organize and retrieve these data efficiently has become to a hotspot at present.We firstly give the state of the art and the trend of development on video classification. Then, based on the analysis of the existed video classification algorithms, we propose a new approach based on multi-features and sequential pattern mining in the dissertation. Such as a three-level model, on the first level, it mainly extracts a set of features which related to the background contexts. Then, we can recognize action event by machine learning. And each video can be present a sequence of event. Finally, we mainly use effective frequent sequential pattern mining algorithm to dispose of the video classification and form a classification rule-lib to match the shot sequence that will be classified. Experimental results demonstrate that the proposed classifying algorithms satisfactory performance in a variety of surveillance videos.