摘要
提出一种基于梯度图像,融合帧间差分和背景差分的运动目标检测新方法。其特点是采用混合高斯背景模型,先利用针对梯度图像的帧间差分检出变化区域,再利用背景差分从变化区域中检出运动物体,最后利用连通性检验消去噪声和阴影。针对真实视频序列的实验结果表明,该方法既简单有效,又具有较小的运算量和较好的鲁棒性。
A novel approach to detection of video moving objects(VMOs) is proposed,which combines grads-based frame difference with background subtraction.Based on a mixed Gauss model of background,it finds firstly changed areas by applying frame difference to grads images derived,detects then VMOs from the areas by using background subtraction,and obtains finally noise-free VMOs by using connectivity test.Proved by the results of experiments on some real video sequences,it is rather simple,effective and robust.
出处
《光电子技术》
CAS
北大核心
2009年第1期34-36,41,共4页
Optoelectronic Technology
基金
国家自然科学基金资助项目(60672026)
关键词
运动目标检测
梯度图像
帧间差分
背景差分
video moving objects detecting
grads images
frame difference
background subtraction
