Abstract:

A new background modeling approach based on object locating is presented in this paper. The method separates the foreground pixels and background pixels preliminary via predicting the location of moving object in video sequences, and then constructs the background model. It effectively overcomes the mixing of background pixels and foreground object in timeaveraging method. Experimental results show that the proposed approach could detect and track the moving object without storing the background model in advance. It makes up for the deficiencies in background subtraction method which needs background model before hand and is lack of adaptation to background changes.

Key words: intelligent surveillance； background modeling； object tracking