基于对象的监控视频摘要生成优化方法

摘要

随着监控技术和网络技术的飞速发展，高清网络监控摄像头广泛应用于各个行业。这些高清摄像头全天候工作，产生了海量的监控视频数据。如何快速完整的浏览长时间的监控视频已经成为监控行业目前亟待解决的问题。视频摘要就是解决“海量视频数据处理”的重要手段。然而，传统的基于关键帧的视频摘要生成方法采用帧采样无法完整表示每个对象的运动轨迹，从而导致大量有用视频信息的丢失。针对监控视频的特点，设计了一种基于对象的视频摘要生成处理框架，对比传统摘要生成方法分析了监控视频摘要生成框架所需的核心技术：目标检测跟踪与轨迹提取、运动目标轨迹的组合优化和轨迹融合之像素融合，得到了一个比原始视频短的多的摘要视频，实现了快速浏览且保留了原始视频中大多数运动对象的信息。%With the rapid development of surveillance technology and network technology,high -definition network surveillance cameras are widely used in various industries.With 24 hours working per day,these cameras capture millions of video.In the field of surveillance industry,fast and completely browsing the long surveillance video become urgent requirement.Video summary generation technology is an effective means to solve this problem.However,a massive of useful video information has been lost because the traditional video summary generation method,based on the key -frame,cannot fully represent the trajectory of each object.This paper focus on surveillance video data and presents an object -based technology framework of video synopsis generation.Compared with the traditional method of generating summary,it analyzes the core technologies of surveillance video summary generation technology framework,i.e.object detection tracking and trajectory extraction,optimization of combined motion target trajectory and pixel fusion,achieves a quick overview and retains most of the original information of moving objects in video.