| As people concern of social public security, Video surveillance system is widely used in all walks of life. Cameras work24hours a day, resulting in a massive video data. How to store the massive video data in a reasonable manner and how to extract and analysis the video content effectively becomes a serious problem. The existing video processing technology, which not only lose the key information of video, also left a lot of spatio-temporal redundancy, cannot be well applied to the video surveillance system. To this end, this paper puts forward video synopsis, based on greedy algorithm and genetic algorithm, which can effectively avoid the above problems.The main idea of video synopsis is to extract the essential activities of the original video, and relax their chronological consistency by shifting them from their original time intervals to other time intervals when no other activities take place at these spatial locations. So the activity in the video is condensed into a shorter period by simultaneously showing multiple activities, even when they originally occurred at different times. Video synopsis realized the deep concentration both in spatio-temporal and content.This process includes four major phases:(ⅰ) the pretreatment of the original video by removing the blank frame using contour-detecting method,(ⅱ) the video structure analysis,(ⅲ) the improvement and optimization of video synopsis using greedy algorithm and genetic algorithm.(ⅳ) the optimization of the system performance using Hadoop.The experimental results by using surveillance video data proved the rationality of the design and effectiveness of the system. |
PDF Full Text Request