大场景视频协同监控系统的研究与实现

Cooperative Video Monitoring System of Large Scene

在"大场景"环境下,采用单个或者多个独立摄像机的视频监控系统无法获得大场景的视觉,导致监控人员不能快速、及时、正确地对特殊情况做出判断,留下了严重的安全隐患。提出了一种大场景视频协同监控系统。该系统采用DirectShow技术进行音/视频采集、FFmpeg技术进行视频压缩、防止磁盘碎片产生的存储方法进行视频存储,引入心跳机制解决图像的实时同步问题。同时,所提出的系统使用了SIFT特征提取、K-D树特征匹配查找、基于RANSAC的精确匹配和双线性加权融合等算法进行图像融合,并通过大场景识别与定位模块提前保存所需要的融合参数,加快图像融合的效率。系统测试表明:所提出的系统的图像融合效果满足在大场景进行视频协同监控的要求,同时平均融合时间耗费也满足实时性监控的要求。

A video system of cooperative monitoring for large scenes is proposed in this paper.DirectShow in audio/ video capturing,FFmpeg compression method,and a method of preventing disk fragmentation generating are used in this system to store video files.To solve the problem of image real-time synchronization heartbeat mechanism is introduced. Meanwhile,in this system,SIFT feature extraction,feature matching K-D tree search,RANSAC matching and bilinear weighted fusion algorithm are used in image fusion,and parameters required for fusion are saved by scene recognition and position- ing module in advance,contribute to accelerating the efficiency of image fusion,