The Research And Implementation Of Video Retrieval System Used For Fire-fighting

With the popularity of multimedia technology, various video-based applications emerged.The traditional retrieval technology which is based on text label can not satisfy the demand of massive video retrieving anymore. How to retrieve desired information from huge video data has become extremely important. Recently, content-based video retrieval mainly concentrated on face reorganization, inanimate object retrieval and received good achievement. This paper mainly focuses on video retrieval aiming at fire-fighting.Recent years, with the investment in fire-fighting increased rapidly in our country, various real-time video capture and restore systems were deployed and gets some achievements.. These bring massive videos captured in the scene of the fire but also introduce a serious problem that is how to retrieve valuable information from such a great amount of video data accumulated in quiet a long time.The content-based video retrieval firstly needs structural treatment on video, which includes shot boundary detection and key frame extraction. This is a key step for implementing a content-based video retrieval (CBVR) system. Secondly, it needs extracting movement features of such frames and save them to video database. Finally, the system should give a series of videos index by relevance according to user demands. This paper focused on the features of the video captured on the sense of fire, designed and realized a video retrieval system based on fire-fighting. The system includes label-based video retrieval and contend-based video retrieval subsystem. Specially, this paper proposed an algorithm for shot boundary detection, which can overcome the interference of shake of the video effectively. What's more, this paper also does some optimization work for the features of video captured by fire sense.This system can help fireman to seek desired information from massive videos accumulated by video capture and restore system.