摘要
针对视觉背景提取(visual background extractor,ViBe)算法中出现的鬼影、阴影前景和漏检静止目标等问题,提出了经过改善的ViBe算法。首先在背景模型初始化阶段,利用改善的平均背景建模法建立初始化背景模型以消除鬼影。在前景检测阶段,通过计算运动目标与背景像素间的差值消除由于光照所造成的阴影,同时利用直方图相似度检测静止目标。最终在车道设置虚拟线圈检测区以检测各个车道的占用比,从而得出道路的交通情况。实验表明,经过改善后的ViBe算法在抑制鬼影出现、检测静止目标和消除阴影三个方面有较好的效果,能够准确检测出道路车流量。
Aiming at the problem of appearing the ghost,the shadow and undetecting stationary target which existed in ViBe algorithm,this paper proposed an improved algorithm.The algorithm utilized improved average background modeling to establish initialization background during initialization,thus suppressed the ghost.The algorithm calculated difference value between the moving object and the background pixel during the process of pixels matching,thus eliminated the shadow.At the same time,the method utilized comparing histogram similarity,thus detected stationary target.Finally,it set virtual detection area in vehicle lane to obtain the road traffic situation.Experiments show that compared to traditional algorithm,the improved algorithm presented perform well in suppressing the ghost,detecting stationary object and eliminating shadow.
作者
黄鑫
肖世德
Huang Xin;Xiao Shide(College of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610036,China)
出处
《计算机应用研究》
CSCD
北大核心
2018年第10期3176-3178,共3页
Application Research of Computers
基金
四川省科技支撑资助项目(2015GZX0101)
四川省应用基础研究基金资助项目(2014JY0212)
关键词
背景差法
车辆检测
鬼影抑制
静止目标
阴影消除
background extractor
vehicle detection
suppressing ghost
stationary object
eliminative shadow
