摘要
视频监控面对的场景具有静态的成分和缓变的成分。为了从视频流中稳定地抽取较精确的背景,提出了一种基于场景区分的背景抽取算法。首先通过混合高斯模型对视频序列中稳定的成分进行建模,然后采用Kalman滤波模型对视频序列中变化的成分进行建模,最后采用场景的区分算法将2种模型结果进行处理从而得到背景。实验结果表明:该方法能从视频流中稳定地抽取相对精确的背景。
The scene monitored by intelligent video is stable most of the time and slow changing temporarily. In order to stably and precisely extract the background from the video stream, a background extract algorithm based on the scene distinction has been presented. The mixture of Gaussians and Kalman filter are employed to process and model the stationary parts and slow changing parts separately, and then the scene discrimination algorithm is used to extract the background by processing the result of those two models. Experimental results indicate that the method can stably extract relative precise background from the video stream.
出处
《传感器与微系统》
CSCD
北大核心
2011年第11期39-42,共4页
Transducer and Microsystem Technologies
基金
国家科技重大专项基金资助项目(2010ZX03006-004)
国家重点基础研究发展计划资助项目(2011CB302906)
关键词
背景抽取
混合高斯模型
卡尔曼滤波器
场景区分
background extraction
mixture model of Gaussians
Kalman filter
scene distinction
