Summarization, Visualization and Browsing of Massive Video Archives

Combat aerial videos are invaluable for tactical military operations, but analyzing and retrieving videos can become a burden as the size of the archive increases. Automated techniques to represent, analyze, and query pertinent video content from vast archives are necessary. We break the problem down into three steps: video pre-processing, summarization at the coarse and finer levels, and visualization and browsing of the query results. In this SBIR Phase I effort, we propose to develop algorithms for visualization, summarization and browsing to allow video analysts to quickly examine and filter large amounts of combat aerial video. Our system will use in-house pre-processing algorithms for video stabilization, detection, and tracking. We will use flow field analysis for sequence segmentation and coarse summarization, and 3D non-rigid factorization analysis for finer summarization of combat videos. The system will provide automated summarization and interactive visualization and browsing of the videos. The outcome of this process will be rapid analysis and browsing of video sequences, enabling analysts to exploit combat videos at will.