摘要
针对智能视频监控提出了一种融合两种运动信息的分级运动检测算法。算法分别在像素级、区域级和帧级进行处理。像素级处理中,运用了改进的基于混合高斯模型的背景相减法,并提出了一个能够及时适应背景快速变化的累积时间差分法,两种方法检测得到两种不同的运动信息,区域级处理融合了这两种运动信息,使算法既能适应背景的缓变,又能适应背景的快变,解决了背景物体运动的问题。帧级处理主要解决全局光照突然变化的问题。实验结果表明,这个算法能够实现稳健可靠的运动检测。
A hierarchical motion detection algorithm, which fuses two types of motion information, is presented in this paper to handle the slow or fast changes of the background and the global illumination changes. The algorithm consists of three distinct levels-pixel level, region level and frame level. At the pixel level, an improved MOG-based background subtraction is introduced to get the better adaptability to the slow change of background. A method of cumulate temporal differencing, which is adaptive to the fast change of the background, is proposed to serve as auxiliary information to improve the result of background subtraction. On the region level, the moving pixels are grouped to get moving region. Then the fusion of background subtraction and cumulate temporal differencing is performed to get real moving foreground. Region level processing provides the solution for the slow and fast changes of the dynamic environments. Finally frame level analysis is performed to detect the global illumination changes and re-initialize the background model or switch among the multi-background models. Results of experiments show the effectiveness and robustness of the algorithm.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2005年第5期552-557,共6页
Pattern Recognition and Artificial Intelligence
关键词
智能视频监控
运动检测
背景模型
Intelligent Video Monitoring, Motion Detection, Background Model
