A method for video synopsis based on Multiple Object Tracking
Bin Zhu, Wen Liu, Gang Wei, Lin Yuan · 2014
Millions of IP camera surveillance city, bank, airport, school and more, especially the increase of surveillance cameras that work 24 hours a day, massive surveillance video data presents formidable challenges to its browsing, retrieval and storage. So we propose to a short video that will be a synopsis of an endless video streams to overcome the problems, and this video synopsis is an effective way to solve this problem to provide a compact video representation, while preserving the essential activities of the original video. Firstly, using Visual Background Extractor (ViBe) motion detector algorithm extracted interest motion blob and frame segmented from the video, and that information compose aspatio-temporal object tubeelement; Secondly, we propose Multiple Object Tracking (MOT) technology match to interest blobs with effective frame build up the object motion tube in the camera surveillance view; Thirdly, using this paper propose the tracker sort rule, that according to the interest object maintain the view time, from long to short array to the tracker vector; Finally, the selected different array vector and pick up objects are stitched to the background image. Some experiments are performed using the proposed algorithm and the results are acceptable.