Abstract:

This paper studies the usual problem of ghosts in moving objects detection based on background subtraction. After analyzing the pixel distribution of    the border area of every foreground blob（PDBA）, a method of ghosts detection is proposed. Firstly, histogram matching（HM） and mean variable ratio（M  MVR） are used to weigh the difference of the PDBA between the current and previous images. Secondly, histogram matching is also used to weigh the diffeerence of the PDBA between every foreground blob and the background connection with that blob. Finally, a threshold segmentation method is used to determine the ghost＇s blob with extremely high HM and low MVR. Experimental results show that the developed algorithm not only detects and removes ghosts f ast and effectively, but also overcomes the shortcomings of other algorithms.

Key words: moving object detection;background subtraction;ghosts