Research On The Algorithm Of Foreground Detection And Background Modeling In Video

As an important branch of computer vision,background modeling and foreground detection is an important research topic in image processing, and it's an important task in Image preprocessing,the purpose is to separate the moving objects from the scene.In intelligent transportation, video surveillance,air detection,it has a wide range of applications. However in the actual scene, due to the influence of noise, illumination of the mutation, background elements, camera jitter or mobile factors, foreground objects from the background separation is still facing a lot of challenges and difficulties.Under this background,this thesis makes a deep research on the detection algorithm of the static and dynamic scenes.The data is collected by the camera fixed at a point of view in the static scene,and the data is collected by the camera fixed at a point,Scene acquisition from multiple perspectives.In this paper, we mainly consider how to separate the moving object from the background, and improve the robustness of the foreground detection in different environments,and how to eliminate the interference of noise and eliminate the interference of background in the dynamic scene.1. A foreground detection algorithm based on algebraic multigrid method and inter-frame difference detection algorithm is proposed,AMG(Algebraic Multi-Grid) method can be used to construct the coarse grid.Result image is constructed using the interpolation of the data of coarse grid, and the layers of coarse grid can be up to three from the experiences. Then the result image be used in the frame difference method to get the target.In order to verify the experimental results,the proposed method is compared with the original method and the comparision.Experimental results show that the method can effectively remove the interference of noise in the static scene,the detection results are better than the original method from the first layer to the third layer,especially in the third layer,removing the interference of noise basically,extract the contour of the moving object effectively.2. On the combination between the algebraic multigrid method and the background difference method for foreground detectionis proposed here,to verify the abilty of algebraic multigrid method to reconstruct the image effectively,in this paper, the effect of reconstruction is applied in the background subtraction method.Experimental results show that the reconstruction results can keep the image informationeffectively,and the moving object can be extracted,and using background subtraction method to detect,it can extract the moving object effectively,the new method has high robustness.3. In the dynamic scene,the paper puts forward a background modeling method based on the panorama,the panorama can be got using image mosaic algorithm and used in the detection of moving objects,in order to establish the background modeling.Based on the panorama, the algebraic multigrid method is used to reconstruct the image,then detecteding the moving object.Experimental results show that it can eliminate the effects of jitter camera and noise on the detection results and get the contour of the moving object based on the panorama.