Research Of Video Summarization And Compression For Intelligent Surveillance

Video surveillance system crowded almost every city, record what happens in every corner of the city. Provide safety for the cities, at the same time produced huge data. How to find useful details and effectively store these video is an urgent problem for intelligent video surveillance system. In this paper, we research motion aware video summarization and compression, which can quickly search the video and effectively store video. The main work we have done as follows:First, we pay attention to video summary and search. Analyzed the mainstream video summary methods and sophisticated video summarization systems, summarized the common and key technologies. For the special use of surveillance videos, we proposed video summary method based on video scenes analysis. Classified the video scenes into three categories, we present three diffident methods for each corresponding videos, programed an application software systems for surveillance video and it works well under actual conditions, lay the foundation for video search. This software has been copyrighted.Second, for video compression.We find that under the static monitoring scene, people only pay attention to moving objects or a region of interest (ROI). Based on this, a method for image cropping (seam carving) has been introduced into motion aware surveillance video compression. We first extract region of interest (ROI), then crop the Non-ROI. In the meantime, take time and space cost of frames into account. Experiments show that, this method not only compressed surveillance video, but also retained the area of interest of videos.