摘要
利用混合高斯模型进行视频车辆检测.在背景建模阶段,易受初始时刻行驶车辆前景的干扰,造成背景模型包含大量前景信息,对后续车辆检测产生消极影响;故而将初始10帧各像素点处的灰度值看作时间域上的序列值,通过格拉布斯准则剔除代表前景信息的异常灰度值,进而对各像素点处代表背景信息的灰度值求取均值和方差,建立背景模型.实验结果表明该方法明显优于传统混合高斯模型背景建模,将其应用于基于混合高斯模型的视频车辆检测系统中,取得了良好的实验效果.
When vehicle detection in video is practiced ,it is easy to be interfered by vehicle foreground of initial time in the background modeling phase ,w hich makes background model have a lot of information of foreground so as to have negative impact on vehicle detection .Therefore ,this paper considered the grey values of each pixel point in initial 10 frames as the sequence values in the time domain to eliminate the ab‐normal grey values that reflected foreground with Grubbs criterion .Then ,background mode was built via calculating mean and variance of grey values that reflected background for each pixel point .The experimen‐tal results showed that the methods put forward by this paper had better performance than GMM back‐ground modeling and was preferably used in video vehicle detection system .
出处
《徐州工程学院学报（自然科学版）》
CAS
2015年第2期15-18,70,共5页
Journal of Xuzhou Institute of Technology(Natural Sciences Edition)
基金
陕西省交通运输厅科研项目(12-26K)
国家山区公路工程技术研究中心开放基金项目(gsgzj-2011-08)
关键词
视频车辆检测
混合高斯模型
背景建模
格拉布斯准则
video vehicle detection system
Gaussian mixture model
background modeling
Grubbs criterion
