交通视频的移动目标检测算法研究  

Research on moving target detection algorithm for traffic video

作　　者：黄金海 梁晓梅[1] 桂邦豪 HUANG Jinhai;LIANG Xiaomei;GUI Banghao(Institute of Information Technology of GUET,Guilin 541004,China)

机构地区：[1]桂林电子科技大学信息科技学院,广西桂林541004

出　　处：《现代电子技术》2019年第5期53-56,共4页Modern Electronics Technique

基　　金：广西壮族自治区教育厅资助项目(KY2016YB815)~~

摘　　要：对视频序列的目标进行识别和跟踪监控是计算机视觉领域的关键技术。依据图像处理和目标检测研究方向的热点算法,将帧间差分与混合高斯背景模型融合法应用于交通视频移动车辆的目标检测分析中。该算法通过背景差分求出光照影响的动态阈值范围,若光照检测阈值小于该动态阈值,选择混合高斯背景模型法,大于该动态阈值则选取帧间差分法。再通过开运算、闭运算、连通域计算,以及区域的凸包形态学运算即可准确描述目标。实验结果表明,该融合算法消除了鬼影现象,在光照环境改善的情况下可以有效应用于交通视频中移动车辆的目标检测。It is a key technology for target identification and tracking monitoring of video sequence in the field of computer vision.A fusion method of inter.frame difference and Gaussian mixture background model is proposed on the basis of hotspot algorithm of image processing and target detection research directions,which is applied to the target tracking analysis of moving vehicle in traffic video.The algorithm can solve the dynamic threshold range influenced by illumination by means of background subtraction.If the light detection threshold is smaller than the illumination dynamic threshold,the Gaussian mixture background model is selected,otherwise the inter.frame difference is selected.The target can be accurately described by the opening operation,closed operation,connected domain calculation,and convex hull morphological operation of the region.The experimental results show that the fusion algorithm can eliminate the ghost phenomenon,and can be effectively applied to the target detection of moving vehicles in traffic video under the condition of illumination improvement.

关 键 词：背景差分 帧间差分 混合高斯模型 移动目标检测 动态阈值 交通视频 

分 类 号：TN911.73.34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]