Abstract: In the case of a fixed camera, background subtraction method is one of the most effective method to extract moving objects in video, and the key of the background subtraction algorithm is the background modeling method and the subtraction method between background and foreground, when in complex background, the background subtraction method may leads to target defect, empty hole and related background modeling methods are complex, which causes long extraction time, so here proposes a multi fusion background subtraction algorithm.By exerting the color characteristics of RGB color model, anti disturbance of codebook method, and rapidity of three frame difference method and Surendra method fully, a new background modeling algorithm and its reduction method are proposed, which reduces the calculation quantity, improves the tolerance of twinkling background, and solves the problem of defect and cavity of moving object in complex background successfully.