Dynamic Background Modeling Based On Stability Analysis

Background modeling is an essential part of moving target detection.In the practical application environment,the background is usually dynamic.Dynamic background refers to that the background is at an unstable state due to the moving of the object.It will affect the detection and reduce the accuracy of detection.It is one of the major challenges of moving target detection.In the dynamic background,the moving of the objects is complex and varied and affected by noise and light.It is very difficult to establish a robust background model.So it makes sense to study the problem of model establishment under dynamic background.Aiming at the problem of high false detection rate of moving target detection in the dynamic background,this paper proposes a dynamic background modeling algorithm based on stability analysis.The main work is as follows:(1)Analyze the relationship between the moving of the dynamic background and the observation window.The dynamic background will show a different state when it is observed under different size window.If one moving object is observed under a certain large window,its statistical properties will tend to be relatively stable.(2)Analyze the relationship between the image down-sampling and the observation window.When the image is down sampled,the resolution of the image reduced,which is equivalent to that,the size of the observation window becomes larger.(3)Gaussian Pyramid down-sampling and Gaussian model is set up respectively in each layer of the Pyramid.(4)Definition of the stability.Each pixel self-adaptively chooses the layer of Gaussian Pyramid that satisfies the stability condition according to its movement characteristics.In the layer,the change of pixel value can be described by a very small number of Gaussian models,and the dynamic background is presented as several large probability events,which is relatively stable.Determine whether the pixel belongs to the background Gaussian distributions of the stable layer.In view of the influence of dynamic background on the moving target detection,this paper proposes a background model algorithm from the perspective of stability.The experimental results demonstrate that the proposed algorithm has a good robustness,and improve the detection precision of the moving target detection under the dynamic background.