An improved Gaussian mixture background model with real-time adjustment of learning rate
LI Ying-hong, Tian Hong-fang, Yan Zhang · 2010
In this paper, an adaptive background modeling approach for moving object detection is proposed. Based on mixture Gaussian model suggested by Stauffer, a mixture Gaussians model has been built for each pixel and its learning rate can be adjusted dynamically according to the scene change from the frame difference. This approach has changed the strategy used in various improvements to re-initialize the model on the condition of light suddenly change. Experiments show that the adaptive background model proposed in this paper has good adaptability to complex environments, the convergence rate of the model can be speeded up, and the moving object can be detected effectively and rapidly in the case of light suddenly changing.