The Achievement Of Video Retrieval System Based On Video Abstract

With the constant development of key video technology including optical technology and image processing technology. HD network monitoring cameras have come into the market and become main video surveillance equipment. And resolution of surveillance video is gradually increased from 480P(PD standard) to 720 P and even 1080P(HD standard), which allows people to see details of scenes and persons captured by a surveillance camera more clearly. While bringing convenience, the HD trend of video surveillance is also producing massive monitoring data. Under such a circumstance, the time cost and economic cost for analysis of video data relying on manpower will be unacceptable to any industry. Therefore, the demand for finding effective ways to enhance video data search capability by means of extracting key video information based on automatic approaches is becoming most urgent in the field of video surveillance.To meet the above mentioned demand, in this paper, combining innovative video retrieval algorithms including video abstraction processing algorithms of visual detection and video tracking and target classification, an innovative video retrieval system based on video abstraction is established. Firstly, the abstract retrieval system abstracts objects containing main video information from surveillance video through target detection and tracking technology. The system then perform characteristic analysis for the main part of captured main video and classify it according to different features of it and create feature index for the convenience of query in the future. Finally, abstract video with retrieval capability is generated through combining the background obtained during background modeling and extracted extracted main part of video by using representative scenes established based on background modeling technology. Through the three steps, namely, detection, classification and composition, redundant data the monitoring people don’t care about can be removed from surveillance video, so as to effectively reduce time required for data analysis of surveillance video and increase query efficiency of monitoring information.This paper firstly introduced domestic and overseas development trend and present situation of surveillance video processing and simply explains key technology of the system. Through analyzing and contrasting between different video processing techniques, technology selection for the system is then conducted according to actual demand. Next, based on the realization of relevant video processing algorithms, detailed design of the overall level of the video abstraction system and specific functional modules is performed. Finally, the realization process of each module is given according to the system design scheme. After the design of the system is finished, its function is tested from test cases of various types of monitoring scenes.