Research Of Surveillance Video Synopsis Algorithm Based On Object

With the development of social economy and computer technology, people on social public security demand is rising. A lot of surveillance cameras are installed to monitor and record the conditions of the public places, so a mass of monitoring video data are produced. When we need to watch and analyse the surveillance video data of the past time, the traditional way like fast forward or backward to browse the surveillance video data will waste much time and has the lower efficiency. Obviously, some important information may be missed. Therefore, how to concisely express the surveillance video data content and convenient to browse, search and analysis of the surveillance video, become a research hotspot and difficult problem of the video monitoring field.In recent years, the traditional surveillance video summarization technology has made a great development. The browse, search and analysis of the surveillance video data become more easily. But the traditional video summarization technology make "frame" as the basic unit, the generation of the video summarization is the extraction of static "key frame" set. But the monitoring cameras in24hours one day produce the amount of surveillance video only on the traditional video summarization and spend a lot of time to analyse the surveillance video.In this paper the video synopsis technology based on object is studied, It have broken the limit that the frame is the basic unit in the traditional video summarization technology. The frame image is changed, so the surveillance video summarization is further shorten the length, and the storage space is greatly saved. And when we need further inquires the process of object’s activity in the original surveillance video, we can use the time stamp of object in surveillance video to directly obtain the appearance and disappearance of time of object in the original surveillance video, then we can search the object in the original surveillance video easily. The main content of this paper include the following aspects:1) Introduce the current main video summarization (synopsis) methods, including the video summarization based on key frames and video synopsis based on object and analyse the classic methods.2) Use a new method to analyse the surveillance video whether it contains activity information. Extract the activity frames from a selection surveillance video, pick up the video clip that include the activity information and filter out that does not include activity information. So the long video is shorted in order to provides the convenience for the after work.3) The objects in the video including activity objects are detected and tracked, then the activity object image is matted by grabcut method4) The proper background image is selected as the video synopsis’s background and the background is updated continuously according to the time change.5) Put the activity object extracted from the video into the background image that has been selected, so the last video synopsis is formed.