Abstract: Visual background extractor model selected each pixel’s spatial neighbourhood pixels randomly to initialize the background model in the first frame,as the result to emerge ghost easily in the early of detection.Meanwhile,the segmentation of background and foreground by global threshold cannot adapt to the dynamic background and the fixed update rate is inappropriate for the scene change rapidly.In order to solve the problem,this paper proposed an enhanced algorithm using pixel’s information in time domain to initialize the background model.At first,this paper took advantage of the history pixels in the image sequences to complete background model initialization.And then,the segmentation threshold was obtained adaptively by the complexity of background using the spatial neighbourhood pixels.Finally,the background model with dynamic updated rate,so that it could adapt to the scene change faster and better.Experimental results show that the enhanced ViBe algorithm not only can remove ghost quickly and effectively,but also can improve the accuracy of the target detection.

Key words: Visual background extractor,Background model,Ghost,Dynamic update