A Novel Method for Background Modeling and Update

This paper proposes a novel background modeling and update algorithm.Firstly a novel Hausdorff-like background modeling is built and for each image sequence foreground is obtained by background subtraction,then a foreground object classifier is used to get moving object,static object,false object,and noise,finally the background model is updated by a foreground object classification based background update algorithm.Extensive experiments results on indoor and outdoor image sequences demonstrate that the proposed system can effectively build a reliable background model and resolve the deadlock problem which results from false object.