摘要
复杂交通环境视频中运动目标的自动检测是智能视频犯罪侦查系统的关键技术之一。本文提出了一种在Hadoop集群上对复杂交通环境视频中的运动目标进行检测的方法——OHMOFD方法,该方法是对帧差法进行改进,有效地克服了传统帧差法检测运动物体时容易出现孔洞的缺点并适合Hadoop集群并行处理。OHMOFD方法在Hadoop集群上实现了一层次并行运动目标检测。实验表明,车辆行人运动目标检测效果较好,检测效率也比运行在PC单机上的串行检测算法效率有明显提高。
Detection of the moving targets in complex traffic surveillance video is a key technology for smart video criminal investigation system.This paper proposes a method OHMOFD for detecting moving objects in complex traffic surveillance video on Hadoop cluster.This method improved the frame difference method.It effectively overcomes the shortcomings of existing holes in a moving target by using the traditional frame difference method.It also adapts to implementation on Hadoop cluster for parallel processing.OHMOFD mechod has been implemented on Hadoop cluster with one layer parallel moving target detections.The experiments show that the result of detection of vehicles and pedestrian is good and the dection efficiency is better than that on a PC.
作者
李振
冯乔生
LI Zhen;FENG Qiao-sheng(School of Information Science and Technology, Yunnan Normal University,Kunming 650000)
出处
《软件》
2017年第11期147-155,共9页
Software
