一种基于无人机视频影像的车流量统计算法  

A Traffic Flow Statistics Algorithm Based on UAV Video Images

作　　者：张冬梅[1] 卢小平[1] 张航 余振宝 苗沛基 ZHANG Dongmei;LU Xiaoping;ZHANG Hang;YU Zhenbao;MIAO Peiji(Key Laboratory of Spatio-temporal Information and Ecological Restoration of Mines,MNR,Henan Polytechnic University,Jiaozuo,Henan 454000,China)

机构地区：[1]河南理工大学自然资源部矿山时空信息与生态修复重点实验室,河南焦作454000

出　　处：《遥感信息》2020年第1期142-146,共5页Remote Sensing Information

基　　金：2016年国家重点研发计划(2016YFC0803103);河南省高校创新团队支持计划(14IRTSTHN026)。

摘　　要：针对城市智能交通管理系统中车流量统计出现的漏检误检问题,提出一种基于无人机(unmanned aerial vehicle,UAV)视频影像的车流量统计双虚拟检测线算法。算法利用数字图像处理技术和计算机视觉对视频流进行处理,首先对无人机获取的视频影像采用均值滤波进行去噪处理,利用改进的多帧平均方法提取出初始背景,通过背景差分法检测出运动目标,然后使用混合高斯背景模型进行背景更新,设置双虚拟检测线并计算二值图像上位于双虚拟线内的连通区域面积、长宽比,统计出实时的车流量。实例验证结果表明,该方法的准确率在非高峰期达到92.94%,高峰期达到91.62%,为城市智能交通管理提供了可靠的数据支持。Aiming at the problem of missed detection and false detection of traffic flow statistics in urban intelligent traffic management system,a double virtual detection line algorithm for vehicle traffic statistics based UAV video image is proposed in this paper.It uses digital image processing and computer vision to deal with streaming video.First,use the mean filter method to remove noise in video frame that UAV obtains,utilize the improved frame average method to extract the initial background,and use the background difference method to detect moving targets.Then,use the Gaussian mixture background model to update background,set up on the double lines of virtual and calculate the binary image located in the connected double virtual line within the area and aspect ratio,as well as calculate the real time traffic flow.The result of example verification shows that the accuracy of this method reaches 92.9%in off-peak period and 91.7%in peak period,which provides reliable data supporting for urban intelligent traffic management.

关 键 词：无人机 视频影像 双虚拟检测线 虚拟线圈 车流量统计 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]