Intelligent Video Abstract System Based On Content

Video abstract system can filter out the redundant video information in surveillance video, and gets more and more attention because of its practical value. This paper describes the general implementation of the video abstract and the latest research results, and chooses the object-based generation method to build the actual video abstract system.In summary, the results of this paper include three points: 1) Using background separation algorithm to generate video abstract. In this paper,a background separation algorithm is used to extract the content of the video. 2) Comparing various background extraction algorithms:Codebook, ViBe (Visual Background Extractor), GMM (Gaussian mixture model). According to the experimental results, GMM with better robustness and stability is selected as the practical algorithm. 3)Archiving a practical value of the video abstract system. The system is divided into three modules: the video summary module, the picture summary module, the intelligent detection module. Practical work includes building a graphical interface of the system, generating a video summary, adding intelligent detection module. In this paper, the actual test and comparison of the three algorithms are given and the advantages and disadvantages of the data are expounded. This paper describes the overall idea of building a video abstract system. We will describe three modules of the system in detail.