交通监控系统中视频运动目标检测算法研究  

Investigation on Video Moving Target Detection Algorithm in Traffic Monitoring System

作　　者：傅赟[1] 王桂丽[1] 周旭廷 侯学鹏 

机构地区：[1]安徽师范大学物理与电子信息学院,安徽芜湖241000

出　　处：《计算机技术与发展》2017年第8期156-158,163,共4页Computer Technology and Development

基　　金：安徽省优秀人才基金(2010SQRL029);安徽师范大学研究生改革研究项目(2015yjg013)

摘　　要：运动目标检测不仅是计算机视觉领域里一项重要的研究内容,也是城市交通监控系统中至关重要的部分,在机器人导航、无人驾驶、医学图像处理以及视频压缩和传输领域都有广泛的运用。在研究光流法、帧间差分法、背景差分法三种目标检测算法原理并对比分析各算法优缺点和适用范围的基础上,在城市交通监控系统中对所选取的同一视频帧分别进行了算法对比仿真实验,并对仿真结果进行了对比分析。仿真实验结果表明,光流法适用于运动状态下的动态目标检测,帧间差分法适用于车速较低的路段,与其他算法相比,背景差分法在城市交通监控系统中的目标检测效果最好,同时也具有运用边缘检测和数学形态学对车辆目标进行标记的能力,可使目标检测更为准确、有效。Detection of moving target not only is a key research content in the field of computer vision, but also plays an important role in the urban traffic monitoring system, which has been widely used in robot navigation, unmanned vehicle, medical image processing, video compression and transmission field and so on, Based on investigation of the principles, advantages and disadvantages, and applications ran- ges of three different kinds of target detection algorithm like the optical flow method, frame difference method, and background difference method, comparative simulation experiments of the algorithms have been conducted with the same video frame selected from city traffic monitoring system as well as comparative analysis on the simulation results. Experiment results have shown that optical flow method is suitable Ibr dynamic target detection under the motion state, and frame difference method can be used for low speed, while the background difference method is the most suitable in city traffic monitoring system of target detection effects compared with other algorithms, and also that edge detection and mathematical morphology tagged on vehicle targets can promote the accuracy and effectiveness of target detec- tion.

关 键 词：城市交通 运动目标检测 光流法 帧间差分法 背景差分法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]