Surveillance Video Information Retrieval System Base On Video Abstract

With the growing technology of monitoring and network, the HD network surveillancecameras have been widely used in various places to meet the security needs of all sectors ofsociety. While the number of cameras is also growing, producing a large number ofsurveillance video, but we do not have so many people and time to deal with this massivevideo data. Intelligent and automatic video surveillance has become a urgent need.To solve such problems, this paper develop a video content search system base on videoabstract. This system provides a method for generate video abstract quickly, and supportsreal-time video or lacal video processing. Video content search is based on the video abstract,it just search the meaningful part of the video, thus we can save a lot of time to search.Video abstract algorithms include three parts: moving object detection, tracking movingobject and generate abstract. Fist we detect the moving objects, and then track each movingobjects, finally traverse every moing events to generate video abstract. Moving objectdetection and tracking moving object are real-time. Video retrieval include five simple featureretrieval: the time when the event happened, the direction of the moving object, the invasionzone, the main color of the object and the type of the object.This article first introduce the processes of video abstract and video retrieval and thetarget of the system. Then give a detail description of algorithm for each process. Finallyanalysis the experimental results of the algorithm, note the defects and the aspects to improveof the system.